\documentclass[12pt]{article}
%\usepackage[latin1]{inputenc}
\usepackage{pdflscape}
 
%\usepackage[onehalfspacing]{setspace}
%\usepackage{aer}
 
\usepackage{natbib}
\usepackage{amsmath}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{graphicx}
\usepackage{url}
\usepackage{rotating}
\usepackage[breaklinks=true]{hyperref}
\usepackage{url}             
\usepackage{tabu}             

\usepackage{pdflscape}
\usepackage{pdfpages}
\usepackage[doublespacing]{setspace}
%\usepackage{comment}
\usepackage{rotating}
%\usepackage{endfloat}
\usepackage{float} 
\usepackage{url}             
 \usepackage{booktabs,caption,fixltx2e}
\usepackage[flushleft]{threeparttable}
 \usepackage{subfiles}
\usepackage{xcolor,colortbl}
\usepackage[]{listings}

 \usepackage{graphicx}
%\graphicspath{{figures/}{../figures/}}
\usepackage{subcaption}


\newtheorem{assumption}{Assumption}
\newtheorem{prediction}{Prediction} 
\newtheorem{theorem}{Theorem}
\newtheorem{algorithm}{Algorithm}
\newtheorem{axiom}{Axiom}
\newtheorem{case}{Case}
\newtheorem{claim}{Claim}
\newtheorem{conclusion}{Conclusion}
\newtheorem{condition}{Condition}
\newtheorem{conjecture}{Conjecture}
\newtheorem{corollary}{Corollary}
\newtheorem{criterion}{Criterion}
\newtheorem{definition}{Definition}
\newtheorem{example}{Example}
\newtheorem{exercise}{Exercise}
\newtheorem{lemma}{Lemma}
\newtheorem{notation}{Notation}
\newtheorem{problem}{Problem}
\newtheorem{proposition}{Proposition}
\newtheorem{remark}{Remark}
\newtheorem{solution}{Solution}
\newtheorem{summary}{Summary}
\usepackage{verbatim}
%\usepackage{ntheorem}
\newtheorem{hyp}{Hypothesis}
 \newenvironment{proof}[1][Proof]{\noindent\textbf{#1.} }{\ \rule{0.5em}{0.5em}}
 
\usepackage{xcolor}
\hypersetup{
    colorlinks,
    linkcolor={red!50!black},
    citecolor={blue!50!black},
    urlcolor={blue!80!black}
}


\usepackage{tabularx}
\newcolumntype{L}[1]{>{\raggedright\arraybackslash}p{#1}} % linksbündig mit Breitenangabe
\newcolumntype{C}[1]{>{\centering\arraybackslash}p{#1}} % zentriert mit Breitenangabe
\newcolumntype{R}[1]{>{\raggedleft\arraybackslash}p{#1}} % rechtsbündig mit Breitenangabe

%\usepackage[detect-all]{siunitx} 			
\def\sym#1{\ifmmode^{#1}\else\(^{#1}\)\fi}	
 
 
\newcommand\startappendixtables{%
    \makeatletter
       \setcounter{table}{0}
   \setcounter{table}{0}\global\def\thetable{A\arabic{table}}%
    \makeatother}
 
\newcommand\startrefereetables{%
    \makeatletter
       \setcounter{table}{0}
   \setcounter{table}{0}\global\def\thetable{R\arabic{table}}%
    \makeatother}
   
 
\newcommand\startappendixfigures{%
    \makeatletter
       \setcounter{figure}{0}
   \setcounter{figure}{0}\global\def\thefigure{A\arabic{figure}}%
    \makeatother}
   
\usepackage[sc]{mathpazo}
\usepackage[margin=1.2in, nohead]{geometry}
%\linespread{1.8}
\setstretch{1.1}


\usepackage{etoolbox}
%\patchcmd{\thanks}{#1}{\protect\singlespacing#1}{}{}
%%squeeze stuff a bit

 \usepackage[compact]{titlesec}
    \titlespacing{\section}{0pt}{2ex}{1ex}
    \titlespacing{\subsection}{0pt}{1ex}{0.25ex}
    \titlespacing{\subsubsection}{0pt}{0.5ex}{0ex}
    \setlength{\parindent}{1.5em}

\usepackage{footmisc}
\renewcommand{\footnotelayout}{\setstretch{1.0}}

\newcommand{\code}[1]{\texttt{#1}}
\def\code#1{\texttt{#1}}
%\usepackage{minted}
 
\begin{document}
\title{Replication Archive for ``Cohesive Institutions and Political Violence''} 
%: Evidence from Nigeria\vspace{-0.5ex}}
\date{\today}
\author{Thiemo Fetzer and Stephan Kyburz}\vspace{-0.5ex}
\maketitle


 
\section{Replication instructions}
This document sets out instructions to replicate all figures and tables in the main paper and the online appendix. There are four main do-files that can be executed back-to-back by running the do file \code{run.all.do}. The do-files call the respective replication data sets and perform the various exercises producing output in latex tabular format in the "tables" and "figures" subfolders. Further, the various files also produce temporary data files. These are stored in the "temporary data files" folder.


\subsection{Main data files}
The main data files that are used are stored in the \code{data files} subfolder.

\begin{enumerate}

\item The data set \code{AB\_OILPRICEl12month.dta} provides oil prices that correspond to time lags of the respective Afrobarometer survey rounds that we use. 

\item \code{ABLGCALLOCATIONS.dta} includes the local government federation account allocations data at a monthly level, that we use in the Afrobarometer micro data analysis (\textit{Source}: Federation Account Allocation Committee FAAC). 

\item \code{afrobarometer\_analysis\_V4.dta} provides all variables used from the Afrobarometer. Included in the data set are the Afrobarometer rounds 1 (1999), 1.5 (2001), 2 (2003), 3 (2005), 3.5 (2007), 4 (2008), and 5 (2012). 

\item \code{EVENTS\_GEOPRECISE.dta} includes geo-located conflict event data points provided by ACLED.

\item \code{FAAC\_allocation\_indices\_2006.dta} provides the FAAC allocation index for each local government area (LGA) (\textit{Source}: Federal Account Allocation Committee). 

\item \code{LGC\_ELEC\_FILLED.dta} provides data on local elections at a monthly level.

\item \code{NGA.PAN.MERGED.L7} provides the full data set. 

\item \code{NGABUDGETOILPRICE.dta} provides yearly oil price data (\textit{Source}: Thomson Reuters Datastream).

\item \code{NGGEOPRECISE.dta} provides the geo-located conflict data (ACLED).

\item \code{NONBOKOHARAM.dta} provides the ACLED conflict data excluding events involving Boko Haram.

\item \code{PERMUTATION.dta} provides the permutation data. 

\item \code{PERMUTATION.STATELEVEL.dta} provides the permutation data set at the state level. 

%%\item \code{CROSS-SECTION-COUNTY.dta} is a set of cross-sectional county-level data and measures used in robustness checks.
%
%%\item \code{VOTINGAGEPOP.dta} provides a measure of the voting age population in a county.
%
%\item \code{ZIPTOFIPS.dta} provides a cross walk of ZIP codes to FIPS county codes that is used for the analysis of Gallup data.  
%
%\item \code{NAICS3EMPSHARES.dta} provides 3-digit NAICS employment shares at the county-level used as control variables.
%
%\item \code{ELECTIONS.dta} provide the election data at the county level providing House, Presidential and Senate election results.
%
%\item \code{cumulative\_2006\_2018.dta} this is the CCES data file which is also available for download on \url{https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi\%3A10.7910/DVN/dII2DB6}.
%
%\item \code{EXPORT-JOB-DEPENDENCE.dta} is from \cite{Brookings2017} measuring the number of jobs dependent on exports in a county.
%
%\item \code{CBP.EMPLOYMENTSHARES.TARIFFS.dta} provides a county-level retaliation tariff exposure measure. This is constructed from data from the County Business Patterns and is used only for robustness check Table A1.
%
%\item \code{IMPACTTARIFF.dta} is the main county-level retaliation tariff exposure measure used in the paper. 
%
%\item \code{SIMULATED\_TARIFFS.dta} provides the main dataset constructing the retaliation tariff exposure measure using simulated retaliation responses constructed according to the algorithm described in Section 4.2 of the paper. There are further data files \code{SIMULATED\_TARIFFS.100.dta} and \code{SIMULATED\_TARIFFS.250.dta}. These are identical but provide only the first 100 or 250 simulated retaliation baskets rather than the 1000 constructed in the main file. The smaller files may be used as otherwise memory issues may emerge when studying the Gallup and the CCES data.
%
%\item \code{SIMULATED\_RCA.dta} provides the revealed comparative advantage measure of the actual as well as the simulated retaliation responses.
%
%\item \code{SIMULATED\_EXPSUPPLYELAST.dta} provides the export supply elasticity for the actual as well as for the simulated retaliation responses.
%
%\item \code{SIMULATED\_US\_SELLERPOWER.dta} provides the market share of US exports for the actual retaliation response in import markets as well as the same measure for all simulated bundles.

\end{enumerate}

\subsection{Replication of Figures and Tables}

There are four main do-files that are called by \code{run.all.do}. Executing the \code{run.do.all} do-file produces all outputs, except some of the maps, which were compiled using ArcGIS. 

\begin{enumerate}
\item \code{prepare.do} 
\item[] This file prepares all the data from raw data sets. 
\item \code{lga-panel.do} 
\item[] This do-file produces the following: Table ~\ref{table:lgc_gsa_mo} (basic conflict results when local governments are appointed), Table ~\ref{table:int_lgc_gsa_mo} (elections and conflict), Table ~\ref{table:conflict-levels-alt} (level effect), Table ~\ref{table:triple-diff-alt} (ethnic groups, elections, and conflict triple diff-in-diff). In addition the do-file also produces the following appendix tables: Table ~\ref{table:withinbetweenfaac} (decomposition of variance in the allocation variables), Table ~\ref{table:oil-producing-oil-fields} (oil-producing states), Table ~\ref{table:allocation-decomposition} (decomposition of allocations), Table ~\ref{table:robustness-election} (election related violence), Table ~\ref{table:democracy-dummy-variance-decomposition} (variance decomposition of local democracy dummy), Table ~\ref{table:residual_var_elections_allocations} and ~\ref{table:residual_var_elections_formula} (residual variations), Table ~\ref{table:int-lgc_gsa_mo-withholding} (disruption in revenue sharing), Table ~\ref{table:lgcelecstmgov} (fixed effects robustness), Table ~\ref{table:int_lgc_gsa_mo-dropbokostates} (robustness Boko Haram), Table ~\ref{table:lgc_extra_alloc_mopost} (extraordinary allocations and conflict), Table ~\ref{table:alternativeconflictdata_post} (alternative conflict data), Table ~\ref{table:alternativedependentvariablespost} (transformations of dependent variable), Table ~\ref{table:alternativetemporalresolutionpost} (different temporal resolutions), Table ~\ref{table:alternativefunctionalformspost} (alternative functional forms), Table ~\ref{table:formulaflexiblecontrolint_post} (formula inputs), Table ~\ref{table:conflict-levels} (non-aligned ethnic groups level effect), and Table ~\ref{table:triple-diff} (non-aligned ethnic groups triple difference). The do-file also produces Panel D in Figure ~\ref{fig:allocations-oilprice-elec}, Figure ~\ref{fig:lgcnonlinearities} (non-linear impact of resource shocks), Figure ~\ref{fig:conflictnearelection} (conflict around elections), Figure ~\ref{fig:lgcnonlinearities_conventional} (non-linear effects), Figure ~\ref{fig:leaveoneout} (leave-one-out validation), Figure ~\ref{fig:permutation-test} (permutation test on allocations), and Figure ~\ref{fig:permutation-statelevel-test} (permutation test on election status).
\item \code{afrobarometer.do}
\item[] This do-file produces all the results using the Afrobarometer micro-data. Specifically, it produces Table ~\ref{table:ab-alloc-victim-12m} (victimization and participation), Table ~\ref{table:ab-approval-trust-corrupt} (perception of local governance), Table ~\ref{table:validation_exclusion_ethnic-group-family} (validation of exclusion), Table ~\ref{table:ab-alloc-grievancesnational-12m} (grievances towards national institutions), and Table ~\ref{table:ab-alloc-grievancesstate-12m} (grievances towards state institutions).
\item \code{codingdiscrepancies.do}
\item[] This do-file, that is also run within the \code{prepare} do-file, describes and codes all the potential deviations from the official federal allocations (see Appendix B.2 for a detailed description). The coding is based on a media analysis using the FACTIVA newspaper database. This file prepares the data for the robustness exercise presented in Table ~\ref{table:int-lgc_gsa_mo-withholding}.
\end{enumerate}

\noindent Additional remarks on figures and tables:
\begin{itemize}
\item Figure ~\ref{fig:NG_map_ACLED_lgc_anyconflict_quartiles} was produced in ArcGIS using the ACLED conflict event data.
\item Panels A and C in Figure ~\ref{fig:allocations-oilprice-elec} were produced in ArcGIS using the allocation and local election data. Panels B and Panel D in Figure ~\ref{fig:allocations-oilprice-elec} can be reproduced with the do-file \code{lga-panel.do}.
\item Table ~\ref{table:faacformula} (revenue allocation formula index) is produced already in the do-file \code{prepare} that processes and prepares the data.
\item Figures ~\ref{fig:ethnic-groups-nigeria} and ~\ref{fig:pop_shares_hausa-mumuye_taraba} are produced in ArcGIS using our calculations based on ethnic information in the Demographic Health Surveys (DHS 1990, 2003, 2008, 2010, 2013).
\end{itemize}

%\begin{enumerate}
%\item \code{county.do}
%\item[] This file produces all county-level results, figures and tables. Specifically, it produces Table 1, A1, A2, A3, A4, A5, A6, A7, A8 and A9 as well as Figures 2, Figure A2 and A3. From line 600 onwards it further constructs measures stored in "temporary data files" folder that are subsequently used in the do-file \code{tradeoff.do} 
%
%\item \code{cces.do} 
%\item \code{gallup.do}
%\item \code{tradeoff.do}
%\end{enumerate}
%

\subsection{Files}
The tex-version of this readme loads all the relevant output files and fragments in the tabular format as they are presented in the paper implying that the output in the readme fully replicates the output and format and structure as is in the paper. For reference, the paths to the fragments and files are also provided here for convenience as follows.

\subsubsection{Figures}
Figures are stored in the figures subfolder. The individual figures in the draft are comprised of the following individual files.

\paragraph{Figure ~\ref{fig:NG_map_ACLED_lgc_anyconflict_quartiles}}
\begin{itemize}
\item \code{NG\_map\_ACLED\_lgc\_anyconflict\_quartiles.pdf}
\end{itemize}
\paragraph{Figure ~\ref{fig:allocations-oilprice-elec}}
\begin{itemize}
\item Panel A: \code{NG\_FAAC\_index\_20062013.pdf}
\item Panel B: \code{statutoryallocations-oilprice.png}
\item Panel C: \code{NG\_Share\_elected\_19992014.pdf}
\item Panel D: \code{elected.png}
\end{itemize}
\paragraph{Figure ~\ref{fig:lgcnonlinearities}}
\begin{itemize}
\item \code{watercolor-appointed.png}
\item \code{watercolor-elected.png}
\end{itemize}

\noindent \textit{Appendix:}

\paragraph{Figure ~\ref{fig:conflictnearelection}}
\begin{itemize}
\item \code{event-nearelection-anyevents-tm-timetoelectime-5.png}
\end{itemize}
\paragraph{Figure ~\ref{fig:lgcnonlinearities_conventional}}
\begin{itemize}
\item \code{quintile-effect-anypolmilitiavsciviliansevts-pre.png}
\item \code{quintile-effect-anypolmilitiavsciviliansevts-post.png}
\end{itemize}
\paragraph{Figure ~\ref{fig:leaveoneout}}
\begin{itemize}
\item Panel A: \code{leave-one-out-main-anyevents.pdf} and \newline \code{leave-one-out-int-anyevents.pdf}
\item Panel B: \code{leave-one-out-iv-main-anyevents.pdf} and \newline \code{leave-one-out-iv-int-anyevents.pdf}
\end{itemize}
\paragraph{Figure ~\ref{fig:permutation-test}}
\begin{itemize}
\item \code{permutation-test-interact-both.pdf}
\end{itemize}
\paragraph{Figure ~\ref{fig:permutation-statelevel-test}}
\begin{itemize}
\item \code{permutation-statelevel-test.pdf}
\end{itemize}
\paragraph{Figure ~\ref{fig:ethnic-groups-nigeria}}
\begin{itemize}
\item \code{NG\_map\_ethnic\_groups\_lga.pdf}
\end{itemize}
\paragraph{Figure ~\ref{fig:pop_shares_hausa-mumuye_taraba}}
\begin{itemize}
\item Panel A: (a): \code{NG\_map\_akwa-ibom\_ethnicities\_Ibibio.pdf} and \newline (b): \code{NG\_map\_akwa-ibom\_ethnicities\_oron.pdf}
\item Panel B: (a): \code{NG\_map\_taraba\_state\_hausa\_fulani\_pop\_share.pdf} and \newline (b): \code{NG\_map\_taraba\_state\_mumuye\_pop\_share.pdf}
\end{itemize}

\subsubsection{Tables}
Tables are stored as fragments providing the point estimates, summary statistics (where applicable) and the standard errors. They are assembled in this readme-document into the Latex tabular layout. For convenience the individual table fragments that are used to construct the tables are mapped here to the table environment.

%Table 1
\paragraph{Table ~\ref{table:lgc_gsa_mo}}
\begin{itemize}
\item \code{fragment-lgc\_gsa\_mo-brent-lgc\_elec\_status-anywidelayout-ols-ID\_lgc-stm}
\item \code{fragment-lgc\_gsa\_mo-brent-lgc\_elec\_status-anywidelayout-iv-ID\_lgc-stm}
\end{itemize}

% Table 2
\paragraph{Table ~\ref{table:int_lgc_gsa_mo}}
\begin{itemize}
\item \code{fragment-int-lgc\_gsa\_mo-brent-lgc\_elec\_status-anywidelayout-ols-idelec-stm}
\item \code{fragment-int-lgc\_gsa\_mo-brent-lgc\_elec\_status-anywidelayout-iv-idelec-stm} 
\end{itemize}

% Table 3
\paragraph{Table ~\ref{table:ab-alloc-victim-12m}}
\begin{itemize}
\item \code{fragment-ab6-victim-ols-yrmo\_survey}
\item \code{fragment-ab12-victim-ols-yrmo\_survey}
\item \code{fragment-ab18-victim-ols-yrmo\_survey}
\end{itemize}

% Table 4
\paragraph{Table ~\ref{table:conflict-levels-alt}}
\begin{itemize}
\item \code{fragment-leveleffectplain-lgc\_excl\_ethngr\_share-lgc\_elec\_status-} \newline \code{anywidelayout-ols-ID\_lgc-tm}
\item \code{fragment-leveleffect-dummified-aboveexc-lgc\_elec\_status-} \newline \code{anywidelayout-ols-ID\_lgc-tm}
\end{itemize}

%Table 5
\paragraph{Table ~\ref{table:triple-diff-alt}}
\begin{itemize}
\item \code{fragment-dummified-sample0-heterogenous-lgc\_gsa\_mo-aboveexc-} \newline \code{anywidelayout-ols-idelec-tm}
\item \code{fragment-dummified-triplediff-lgc\_gsa\_mo--lgc\_elec\_status-}\newline \code{anywidelayout-ols-idelec-tm}
\end{itemize}

% Table 6
\paragraph{Table ~\ref{table:ab-approval-trust-corrupt}}
\begin{itemize}
\item \code{fragment-ab6-grievances-ols-yrmo\_survey}
\item \code{fragment-ab12-grievances-ols-yrmo\_survey}
\item \code{fragment-ab18-grievances-ols-yrmo\_survey}
\end{itemize}

\noindent \textit{Appendix:}

% Table A2
\paragraph{Table ~\ref{table:faacformula}}: no fragments

% Table A2
\paragraph{Table ~\ref{table:withinbetweenfaac}}: no fragments

% Table A3
\paragraph{Table ~\ref{table:oil-producing-oil-fields}}
\begin{itemize}
\item \code{fragment-sample0-heterogenous-lgc\_gsa\_mo-oilproducing-} \newline \code{anywidelayout-ols-idelec-stm}
\item \code{fragment-sample0-heterogenous-lgc\_gsa\_mo-oilgasfielddummy-} \newline \code{anywidelayout-ols-idelec-stm}
\end{itemize}

% Table A4
\paragraph{Table ~\ref{table:allocation-decomposition}}
\begin{itemize}
\item \code{allocation-decomposition-brent}
\item \code{allocation-decomposition-crudeNGA} 
\end{itemize}

% Table A5
\paragraph{Table ~\ref{table:robustness-election}}
\begin{itemize}
\item \code{fragment-int-lgc\_gsa\_mo-brent-lgc\_elec\_status-} \newline \code{anywidelayout-ols-idelec-stm-sampleelectionrestricted}
\item \code{fragment-int-lgc\_gsa\_mo-brent-lgc\_elec\_status-} \newline \code{anywidelayout-ols-idelec-stm-sampleelectionrestricted2}
\item \code{fragment-int-lgc\_gsa\_mo-brent-lgc\_elec\_status-} \newline \code{anywidelayout-ols-idelec-stm-sampleelectionrestricted3}
\end{itemize}

% Table A6
\paragraph{Table ~\ref{table:democracy-dummy-variance-decomposition}}
\begin{itemize}
\item \code{democracy-dummy-variance-decomposition}
\end{itemize}

% Table A7
\paragraph{Table ~\ref{table:residual_var_elections_allocations}}
\begin{itemize}
\item \code{decomposition-varsexplore1-lgc\_elec\_status-stm}
\end{itemize}

% Table A8
\paragraph{Table ~\ref{table:residual_var_elections_formula}}
\begin{itemize}
\item \code{decomposition-varsexplore2-lgc\_elec\_status-stm} 
\end{itemize}

% Table A9
\paragraph{Table ~\ref{table:int-lgc_gsa_mo-withholding}}
\begin{itemize}
\item \code{fragment-rob-withholding-month-int-anypoliticalmilitiaevents-lgc\_gsa\_mo}
\item \code{fragment-rob-withholding-int-anypoliticalmilitiaevents-lgc\_gsa\_mo} 
\item \code{fragment-rob-withholdingmax-int-anypoliticalmilitiaevents-lgc\_gsa\_mo} 
\end{itemize}

% Table A10
\paragraph{Table ~\ref{table:lgcelecstmgov}}
\begin{itemize}
\item \code{fragment-lgc\_gsa\_mo-lgc\_elec\_status-anywidelayout-lgcelecstmgov-tm}
\item \code{fragment-int-lgc\_gsa\_mo-tm-lgc\_elec\_status-anywidelayout-lgcelecstmgov-tm}
\end{itemize}

% Table A11
\paragraph{Table ~\ref{table:int_lgc_gsa_mo-dropbokostates}}
\begin{itemize}
\item \code{fragment-int-lgc\_alloc\_mo-brent-lgc\_elec\_status-} \newline \code{anywidelayout-ols-ID\_lgc-stm-bokoharamstate}
\item \code{fragment-int-lgc\_alloc\_mo-brent-lgc\_elec\_status-} \newline \code{anywidelayout-ols-ID\_lgc-stm-bokoharamstatebroad} 
\end{itemize}

% Table A12
\paragraph{Table ~\ref{table:lgc_extra_alloc_mopost}}
\begin{itemize}
\item \code{fragment-int-lgc\_extra\_alloc\_mo-brent-lgc\_elec\_status-}
\newline \code{anywidelayout-ols-idelec-stm}
\item \code{fragment-int-lgc\_extra\_alloc\_mo-brent-lgc\_elec\_status-}\newline \code{anywidelayout-iv-idelec-stm}
\end{itemize}

% Table A13
\paragraph{Table ~\ref{table:alternativeconflictdata_post}}
\begin{itemize}
\item \code{fragment-int-lgc\_gsa\_mo-tm-lgc\_elec\_status-anyotherdatalayout-tm}
\item \code{fragment-int-lgc\_gsa\_mo-tm-lgc\_elec\_status-levotherdatalayout-tm}
\end{itemize}

% Table A14
\paragraph{Table ~\ref{table:alternativedependentvariablespost}}
\begin{itemize}
\item \code{fragment-int-lgc\_gsa\_mo-year-lgc\_elec\_status-anywidelayout-ols}
\item \code{fragment-int-lgc\_gsa\_mo-year-lgc\_elec\_status-levelwidelayout-ols}
\item \code{fragment-int-lgc\_gsa\_mo-year-lgc\_elec\_status-logpcwidelayout-ols} 
\item \code{fragment-int-lgc\_gsa\_mo-year-lgc\_elec\_status-pcwidelayout-ols} 
\end{itemize}

% Table A15
\paragraph{Table ~\ref{table:alternativetemporalresolutionpost}}
\begin{itemize}
\item \code{fragment-int-lgc\_gsa\_mo-year-lgc\_elec\_status-anywidelayout-ols}
\item \code{fragment-int-lgc\_gsa\_mo-tq-lgc\_elec\_status-anywidelayout-ols}
\item \code{fragment-int-lgc\_gsa\_mo-tm-lgc\_elec\_status-anywidelayout-ols}
\end{itemize}

% Table A16
\paragraph{Table ~\ref{table:alternativefunctionalformspost}}
\begin{itemize}
\item \code{fragment-int-reghdfe-levelwidelayout-lgc\_gsa\_mo}
\item \code{fragment-int-ppmlhdfe-levelwidelayout-lgc\_gsa\_mo}
\item \code{fragment-int-xtnbreg-levelwidelayout-lgc\_gsa\_mo}
\end{itemize}

% Table A17
\paragraph{Table ~\ref{table:formulaflexiblecontrolint_post}}
\begin{itemize}
\item \code{fragment-robqq\_population\_wei\_tm-int-lgc\_gsa\_mo-brent-}
\newline \code{lgc\_elec\_status-anywidelayout-ols-idelec-stm}
\item \code{fragment-robqq\_landmass\_weigh\_tm-int-lgc\_gsa\_mo-brent-}
\newline \code{lgc\_elec\_status-anywidelayout-ols-idelec-stm}
\item \code{fragment-robqq\_primary\_enroll\_tm-int-lgc\_gsa\_mo-brent-} \newline \code{lgc\_elec\_status-anywidelayout-ols-idelec-stm}
\end{itemize}

% Table A18
\paragraph{Table ~\ref{table:validation_exclusion_ethnic-group-family}}
\begin{itemize}
\item \code{fragment-ab-SometimesUnfairly-ethnicgrievances-AB\_excluded\_ethn-ols}
\item \code{fragment-ab-AB\_ethngr\_treated\_unfair-ethnicgrievances-AB\_excluded\_ethn-ols}
\end{itemize}

% Table A19
\paragraph{Table ~\ref{table:ab-alloc-grievancesnational-12m}}
\begin{itemize}
\item \code{fragment-ab6-grievancesnational-ols-yrmo\_survey}
\item \code{fragment-ab12-grievancesnational-ols-yrmo\_survey}
\item \code{fragment-ab18-grievancesnational-ols-yrmo\_survey} 
\end{itemize}

% Table A20
\paragraph{Table ~\ref{table:ab-alloc-grievancesstate-12m}}
\begin{itemize}
\item \code{fragment-ab6-grievancesstate-ols-yrmo\_survey}
\item \code{fragment-ab12-grievancesstate-ols-yrmo\_survey} 
\item \code{fragment-ab18-grievancesstate-ols-yrmo\_survey} 
\end{itemize}

% Table A21
\paragraph{Table ~\ref{table:conflict-levels}}
\begin{itemize}
\item \code{fragment-leveleffectplain-lgc\_excl\_ethngr\_share-lgc\_elec\_status-} \newline \code{anywidelayout-ols-ID\_lgc-tm}
\item \code{fragment-leveleffect-dummified-aboveexc-lgc\_elec\_status-} \newline \code{anywidelayout-ols-ID\_lgc-tm} 
\end{itemize}

% Table A22
\paragraph{Table ~\ref{table:triple-diff}}
\begin{itemize}
\item \code{fragment-sample0-heterogenous-lgc\_gsa\_mo-lgc\_excl\_ethngr\_share-} \newline \code{anywidelayout-ols-idelec-tm}
\item \code{fragment-triplediff-lgc\_gsa\_mo-lgc\_excl\_ethngr\_share-lgc\_elec\_status-} \newline \code{anywidelayout-ols-idelec-tm} 
\end{itemize}

%\paragraph{Table ~\ref{}}
%\begin{itemize}
%\item 
%\end{itemize}

\clearpage

\section{Main Results}
The main results presented in the tables and figures below can be replicated directly by running the do-file \code{lga-panel.do}, as described above. Before this do-file can be executed, the data is read in and prepared running the do-file \code{prepare.do}. This do-file calls the main data set \code{NGA.PAN.MERGED.L7.dta}. This \code{lga-panel.do} do-file then estimates the main regressions and produces the bulk of the tables and figures (see above for further detail). The do file has the individual code segments ordered and labelled with the corresponding table and figure numbers as they appear in the text. The same results are obtained just running the do-file \code{run.do.all}.

%Note that you have to adjust the path at the start of the do-file. 

%\bibliographystyle{chicago}
% \setstretch{.25}
%\bibliography{library.bib}
%\bibliography{library}
%\input{MAIN_Nigeria_Fetzer_Kyburz_29052018.bbl}
%\begin{comment}
\newpage
 
 \setstretch{.9}


 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%\section*{Figures and Tables}
\section*{Tables and Figures for Main Text}

%Figure 1: Conflict intensity across Nigeria
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{figure}[h!]
\centering
\caption{Conflict intensity across Nigeria}
\includegraphics[width=0.7\textwidth]{figures/NG_map_ACLED_lgc_anyconflict_quartiles_BW.pdf}
\label{fig:NG_map_ACLED_lgc_anyconflict_quartiles}
\vspace{0.2cm}
\begin{minipage}{0.975\textwidth}
\scriptsize{\textbf{Notes:} Map shows the distribution of conflict events across local government areas (LGA) in the period 1999 to 2014. \textit{Sources}: conflict data is from \textit{ACLED} and administrative boundaries are from Global Administrative Areas (GADM; \href{https://gadm.org/}{\color{black}https://gadm.org/}). }
\end{minipage}
\end{figure}
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 
 \setstretch{.9}

 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{landscape}
\begin{figure}[htp]
\vspace{-0.8cm}
\centering
\caption{Spatial and temporal variation in revenue sharing and the election status across in Nigeria from 1999-2014 \label{fig:allocations-oilprice-elec}}
\begin{array}{ll}
\centering{
\multicolumn{2}{l}{\textbf{Spatial and temporal variation in revenue sharing}} \\
 \text{\textbf{Panel A}: Formula weights} &  \text{\textbf{Panel B} : Allocations}  \\
\includegraphics[scale=.34]{figures/NG_FAAC_index_20062013_BW.pdf} & \includegraphics[scale=0.27]{figures/statutoryallocations-oilprice_BW.png} \\
%&  \includegraphics[scale=0.3]{figures/allocations-oilprice.png} \\
%\text{Panel C: Statutory allocations} &  \text{Panel D: Extraordinary allocations}  \\
% \includegraphics[scale=0.3]{figures/statutoryallocations-oilprice.png} &
 %\includegraphics[scale=0.3]{figures/extraallocations-oilprice.png} 
\vspace{0.1cm}
 \multicolumn{2}{l}{\textbf{Spatial and temporal variation in election status}} \\
  \text{\textbf{Panel C}: Spatial variation in election status} &  \text{Panel B: Temporal variation}  \\
\includegraphics[scale=.34]{figures/NG_Share_elected_19992014_BW.pdf} & \includegraphics[scale=.25]{figures/elected_BW.png}
\\

\end{array}

\vspace{-0.2cm}

\begin{minipage}{1.4\textwidth}
\scriptsize{\scriptsize{\textbf{Notes:} Figure presents the spatial and temporal variation exploited in the paper. The top panels focus on the revenue sharing with \textit{Panel A} presenting the share of revenues that accrue to each local government area (LGA) based on the horizontal revenue sharing formula (Federation Allocation Account Committee FAAC), while \textit{Panel B} presents the monthly variation in overall revenue allocations made to local governments in aggregate (right scale) along with the monthly prices of Brent Crude oil (left scale). The bottom panel focuses on the election status of local government councils (LGC) over time with \textit{Panel C} presenting the variation across space and \textit{Panel D} presenting the variation across all LGCs over time. 
}
\end{minipage}


}
\end{figure}
\end{landscape}

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% FIGURE: POSITIVE AND NEGATIVE SHOCKS
\begin{landscape}
\begin{figure}[h]
\caption{Positive and Negative Resource Shocks Under Elected and Appointed Regime\label{fig:lgcnonlinearities}}
\begin{center}
% \emph{Panel A:} Events with violence against civilians
$\begin{array}{cc}
  %\includegraphics[scale=.3]<1>{figures/quintile-effect-anypoliticalmilitiaevents-pre.png} &  \includegraphics[scale=.3]<1>{figures/quintile-effect-anypoliticalmilitiaevents-post}  \\
%  \includegraphics[scale=.3]<2>{figures/quintile-effect-anymilitaryvspolmilitiaevts-pre.png} &  \includegraphics[scale=.3]<2>{figures/quintile-effect-anymilitaryvspolmilitiaevts-post}  \\
 % \includegraphics[scale=.4]{figures/quintile-effect-anypolmilitiavsciviliansevts-pre.png} &  \includegraphics[scale=.4]{figures/quintile-effect-anypolmilitiavsciviliansevts-post}  \\
  
    \includegraphics[scale=.15]{figures/watercolor-appointed_BW.png} &  \includegraphics[scale=.15]{figures/watercolor-elected_BW.png}  \\


\mbox{(a): Appointed Local Council} &  \mbox{(b): Elected Local Council}\\[3pt]
  \end{array}$

  \end{center}
 \scriptsize{\textbf{Notes}: Figure showcases the non-linear impact of positive versus negative revenue shocks on conflict. The figure presents results from bootstrapped lowess regressions on the residuals of the dependent variable, after having demeaned the data by local government area (LGA) and state-by-time fixed effects. The method first computes lowess regressions from 1000 bootstrapped samples of the demeaned data. It then calculates density estimates of the predictions from the lowess regressions for several hundred cuts along the y-axis and distributes a specified greyscale proportional to that density estimate. The resulting figure displays the uncertainty in the regressions visually. The median value of the bootstrapped lowess predictions is indicated as a solid white line, while OLS regressions are indicated as a black line. Figure \textit{(a)} presents the results for periods with appointed local government councils, while Figure \textit{(b)} displays the results for periods with elected local government councils.}
\end{figure}

\end{landscape}
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% TABLE: MAIN RESULTS RESOURCE RENTS AND CONFLICT (appointed)
\begin{landscape}
 \begin{table}[h!]
 \scalebox{0.8}{
\centering{
 \begin{threeparttable}
 \caption{The effect of resource rents on political violence when local councils are \emph{appointed} \label{table:lgc_gsa_mo}}
\begin{tabular}{l*{10}{c}}
\toprule

                &     &\multicolumn{3}{c}{Type of Event }         &\multicolumn{4}{c}{Groups involved}                      &\multicolumn{2}{c}{Between Pol. Militias \&...} \\
                    
                    \cmidrule(lr){3-5}    \cmidrule(lr){6-9} \cmidrule(lr){10-11}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)} &\multicolumn{1}{c}{(9)}&\multicolumn{1}{c}{(10)}\\
                    &\multicolumn{1}{c}{Overall} & \multicolumn{1}{c}{Battle} &\multicolumn{1}{c}{Civilian Violence}&\multicolumn{1}{c}{Protest}&\multicolumn{1}{c}{Military}& \multicolumn{1}{c}{Pol. Militia}&\multicolumn{1}{c}{Comm. Militia}&\multicolumn{1}{c}{Rebels} &\multicolumn{1}{c}{Military}&\multicolumn{1}{c}{Civilians}\\
\midrule
\emph{Panel A:} OLS \\
\input{tables/fragment-lgc_gsa_mo-brent-lgc_elec_status-anywidelayout-ols-ID_lgc-stm}
\addlinespace
\addlinespace
\emph{Panel B:} IV \\
\input{tables/fragment-lgc_gsa_mo-brent-lgc_elec_status-anywidelayout-iv-ID_lgc-stm}

\addlinespace
\addlinespace

\bottomrule
\end{tabular}

    \begin{tablenotes} 
{\footnotesize \item \textbf{Notes}: Table presents regression results capturing the impact of revenue shocks on conflict incidence focusing on the subsample in which local government councils (LGC) are appointed. All dependent variables are binary indicators capturing whether an event occurred in a local government area (LGA).  All regressions control for state by time fixed effects and LGA fixed effects. The dependent variable includes (1) any violent conflict event defined as the use of force by a group with a political purpose. \textit{Type of events} include (2) battles defined as violent interactions between two politically organized armed groups; (3) violence involving civilians are violent acts by an organized political group against unarmed non-combatants; (4) protests are demonstrations and spontaneous acts of violence by disorganised groups. Columns (5)--(8) include the specific \textit{groups involved}; (6) political militias are armed agents to influence political processes; (7) communal militias are (ethnic or religious) groups engaged in local political competition; (8) rebels are political organizations to counter an established national government. Columns (9)--(10) present results for the specific \textit{dyadic interaction of actors involved}. The main explanatory variable is the monthly revenue allocation to a LGC. The instrumental variable estimation in Panel (B) uses the index weight interacted with the oil price as instrument. Standard errors in parentheses are adjusted for two way clustering by LGA and time with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$.  } 
\end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
\end{landscape}

% OLD NOTES: The dependent variable in column (1) includes any violent conflict event defined as the use of force by a group with a political purpose or motivation. The type of events include battles in column (2) defined as a violent interaction between two politically organized armed groups; violence involving civilians in column (3) defined as deliberate violent acts perpetrated by an organized political group against unarmed non-combatants; protests in column (4) includes demonstrations against a (typically) political entity, involving either protesters or rioters, and spontaneous acts of violence by disorganised groups. Columns (5)--(8) present results for the specific groups involved; political militias in column (6) are defined as armed agents supported by political elites of various types, seeking to influence political processes but not change the government; communal militias in column (7) are defined as groups engaged in local political competition, often traditionally based contests between ethnic, community or local religious groups; rebels in column (8) are defined as political organizations whose goal is to counter an established national governing regime by violent acts. Columns (9)--(10) present the results for the specific dyadic interaction of actors involved in the violence.
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% TABLE: MAIN RESULTS 2: CONFLICT AND ELECTIONS
\begin{landscape}
 \begin{table}[h!]
 \scalebox{0.72}{
\centering{
  \begin{threeparttable}
\caption{Resource rents and conflict -- moderating effect of having an elected local government\label{table:int_lgc_gsa_mo}}
{
\begin{tabular}{l*{10}{c}}
\toprule
                &     &\multicolumn{3}{c}{Type of Event }         &\multicolumn{4}{c}{Groups involved}                      &\multicolumn{2}{c}{Between Pol. Militias \&...} \\
                    
                    \cmidrule(lr){3-5}    \cmidrule(lr){6-9} \cmidrule(lr){10-11}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)} &\multicolumn{1}{c}{(9)}&\multicolumn{1}{c}{(10)}\\
                    &\multicolumn{1}{c}{Overall} & \multicolumn{1}{c}{Battle} &\multicolumn{1}{c}{Civilian Violence}&\multicolumn{1}{c}{Protest}&\multicolumn{1}{c}{Military}& \multicolumn{1}{c}{Pol. Militia}&\multicolumn{1}{c}{Comm. Militia}&\multicolumn{1}{c}{Rebels} &\multicolumn{1}{c}{Military}&\multicolumn{1}{c}{Civilians}\\
\midrule
\emph{Panel A:} OLS \\
\input{tables/fragment-int-lgc_gsa_mo-brent-lgc_elec_status-anywidelayout-ols-idelec-stm.tex}
\addlinespace
\addlinespace
\emph{Panel B:} IV \\
\input{tables/fragment-int-lgc_gsa_mo-brent-lgc_elec_status-anywidelayout-iv-idelec-stm.tex}

\addlinespace
\addlinespace

\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table presents regression results capturing the impact of revenue shocks on conflict incidence during periods when local government councils (LGC) are elected or appointed. All dependent variables are binary indicators capturing whether an event occurred in a local government area (LGA).  All regressions control for state by time fixed effects and LGA by elected status fixed effects. The dependent variable includes (1) any violent conflict event defined as the use of force by a group with a political purpose. \textit{Type of events} include (2) battles defined as violent interactions between two politically organized armed groups; (3) violence involving civilians are violent acts by an organized political group against unarmed non-combatants; (4) protests are demonstrations and spontaneous acts of violence by disorganised groups. Columns (5)--(8) include the specific \textit{groups involved}; (6) political militias are armed agents to influence political processes; (7) communal militias are (ethnic or religious) groups engaged in local political competition; (8) rebels are political organizations to counter an established national government. Columns (9)--(10) present results for the specific \textit{dyadic interaction of actors involved}. The main explanatory variable is the monthly revenue allocation to a LGC. The variable \textit{Elected} indicates whether an LGC is elected or appointed in a given month. The instrumental variable estimation in Panel (B) uses the index weight interacted with the oil price as instrument. Standard errors in parentheses are adjusted for two way clustering by LGA and time with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$.  
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
\end{landscape}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%




%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% INDIVIDUAL LEVEL MAIN RESULTS: VICTIMISATION AND PARTICIPATION
% Table 3
 \begin{table}[h!]
 \caption{Resource rents, elected (vs appointed) local governments and individual level victimization and participation in conflict \label{table:ab-alloc-victim-12m}}
 \scalebox{0.7}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{6}{c}}
\toprule
                    &\multicolumn{2}{c}{Fear of political violence}                    &\multicolumn{2}{c}{Physically attacked}     &\multicolumn{2}{c}{Engage in violence} \\\cmidrule(lr){2-3}\cmidrule(lr){4-5}\cmidrule(lr){6-7}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)}\\
\midrule
\emph{Panel A:}    \\
\input{tables/fragment-ab6-victim-ols-yrmo_survey.tex}
\addlinespace
\addlinespace
\emph{Panel B:}    \\
\input{tables/fragment-ab12-victim-ols-yrmo_survey.tex}
\addlinespace
\addlinespace
\emph{Panel C:}    \\
\input{tables/fragment-ab18-victim-ols-yrmo_survey.tex}

Respondent controls        &          &        X   &          &     X  &           &       X  \\

\bottomrule
\end{tabular}

} 
    \begin{tablenotes} {\footnotesize
\item \textbf{Notes}: Table presents regression results using individual level data from the Afrobaromter survey. All regressions control for time fixed effects and local government area (LGA) fixed effects. The dependent variables in column (1)--(2) indicates how much a respondent personally fears to become a victim of political intimidation or violence; columns (3)--(4) indicates how often the respondent or someone in the respondent's family has been physically attacked in the past year; columns (5)--(6) indicates how often a respondent has used force or violence for a political cause. Panel (A) uses the sum of monthly revenue allocations in the last 6 months, Panel (B) in the last 12 months, and Panel (C) in the last 18 months. The variable \textit{Elected} indicates whether a local government council (LGC) is elected or appointed in a given month. Respondent controls include the respondents age, educational attainment, employment status, gender and an indicator whether the household lives in an urban area. Standard errors in parentheses are clustered by LGA with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$.  
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%    MECHANISMS  %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table 4

\begin{landscape}
  \begin{table}[h!]
  \caption{Inclusion of Ethnic Groups Through Elections and Conflict: Estimation of effect on conflict levels \label{table:conflict-levels-alt}}
 \scalebox{0.67}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{10}{c}}
\toprule
                &     &\multicolumn{3}{c}{Type of Event }         &\multicolumn{4}{c}{Groups involved}                      &\multicolumn{2}{c}{Between Pol. Militias \&...} \\
                    
                    \cmidrule(lr){3-5}    \cmidrule(lr){6-9} \cmidrule(lr){10-11}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)} &\multicolumn{1}{c}{(9)}&\multicolumn{1}{c}{(10)}\\
                    &\multicolumn{1}{c}{Overall} & \multicolumn{1}{c}{Battle} &\multicolumn{1}{c}{Civilian Violence}&\multicolumn{1}{c}{Protest}&\multicolumn{1}{c}{Military}& \multicolumn{1}{c}{Pol. Militia}&\multicolumn{1}{c}{Comm. Militia}&\multicolumn{1}{c}{Rebels} &\multicolumn{1}{c}{Military}&\multicolumn{1}{c}{Civilians}\\
\midrule
\emph{Panel A:} Level effect  \\
\input{tables/fragment-leveleffectplain-lgc_excl_ethngr_share-lgc_elec_status-anywidelayout-ols-ID_lgc-tm.tex}
\addlinespace
\addlinespace
\multicolumn{3}{l}{\emph{Panel B:} Ethnic alignment heterogenous effect}  \\
\input{tables/fragment-leveleffect-dummified-aboveexc-lgc_elec_status-anywidelayout-ols-ID_lgc-tm.tex}
\addlinespace
\addlinespace

\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table presents regression results documenting that local government areas (LGA) with elected local government councils (LGC) experience lower conflict levels and that this effect is driven by places that may have seen significant political exclusion during the periods in which LGCs were appointed. All dependent variables are binary indicators capturing whether an event occurred in an LGA. All regressions control for LGA fixed effects and time fixed effects. The dependent variable includes (1) any violent conflict event defined as the use of force by a group with a political purpose. \textit{Type of events} include (2) battles defined as violent interactions between two politically organized armed groups; (3) violence involving civilians are violent acts by an organized political group against unarmed non-combatants; (4) protests are demonstrations and spontaneous acts of violence by disorganised groups. Columns (5)--(8) include the specific \textit{groups involved}; (6) political militias are armed agents to influence political processes; (7) communal militias are (ethnic or religious) groups engaged in local political competition; (8) rebels are political organizations to counter an established national government. Columns (9)--(10) present results for the specific \textit{dyadic interaction of actors involved}. The main explanatory variable \textit{Elected} indicates whether an LGC is elected or appointed in a given month. The variable \textit{Non-aligned ethnic group} measures the population share that is of a different ethnic group than the state governor. Standard errors in parentheses are adjusted for two way clustering by LGA and time with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$. 
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
\end{landscape}
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table 5: Inclusion of Ethnic Groups Through Elections and Conflict: Triple Difference-in-Differences with Allocated Rents
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{landscape}
  \begin{table}[h!]
  \caption{Inclusion of Ethnic Groups Through Elections and Conflict: Triple Difference-in-Differences with Allocated Rents \label{table:triple-diff-alt}}
 \scalebox{0.67}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{10}{c}}
\toprule
                &     &\multicolumn{3}{c}{Type of Event }         &\multicolumn{4}{c}{Groups involved}                      &\multicolumn{2}{c}{Between Pol. Militias \&...} \\
                    
                    \cmidrule(lr){3-5}    \cmidrule(lr){6-9} \cmidrule(lr){10-11}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)} &\multicolumn{1}{c}{(9)}&\multicolumn{1}{c}{(10)}\\
                    &\multicolumn{1}{c}{Overall} & \multicolumn{1}{c}{Battle} &\multicolumn{1}{c}{Civilian Violence}&\multicolumn{1}{c}{Protest}&\multicolumn{1}{c}{Military}& \multicolumn{1}{c}{Pol. Militia}&\multicolumn{1}{c}{Comm. Militia}&\multicolumn{1}{c}{Rebels} &\multicolumn{1}{c}{Military}&\multicolumn{1}{c}{Civilians}\\
\midrule
\addlinespace
\multicolumn{5}{l}{\emph{Panel A:} Impact of allocations on conflict with appointed LGAs and and Exclusion }\\ 
\addlinespace
LGC Statutory Allocations $\times$  \\
\input{tables/fragment-dummified-sample0-heterogenous-lgc_gsa_mo-aboveexc-anywidelayout-ols-idelec-tm.tex}
\addlinespace
\addlinespace


\addlinespace
\addlinespace

\addlinespace
\multicolumn{5}{l}{\emph{Panel B:} Impact of allocations with elected vs. appointed LGAs and differential exclusion}\\ 
\addlinespace
LGC Statutory Allocations $\times$  \\
\addlinespace
\input{tables/fragment-dummified-triplediff-lgc_gsa_mo--lgc_elec_status-anywidelayout-ols-idelec-tm.tex}


\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table presents regression results documenting that the impact of revenue shocks driving conflict is most pronounced in periods and places were local government councils (LGC) are appointed and where there may be significant share of population excluded (Panel A). Panel B documents that these places are seeing the most drastic weakening of the revenue shocks and conflict relationship. The variable \textit{Elected} indicates whether an LGC is elected or appointed in a given month. The variable \textit{Non-aligned ethnic group} measures the population share that is of a different ethnic group than the state governor. \textit{LGC Allocations} captures the monthly revenue allocation to an LGC. All dependent variables are binary indicators capturing whether an event occurred in a local government area (LGA). All regressions control for LGA by election status fixed effects and time fixed effects. Panel A focuses on the sample with appointed local governments, while Panel B studies the full sample. The dependent variable includes (1) any violent conflict event defined as the use of force by a group with a political purpose. \textit{Type of events} include (2) battles defined as violent interactions between two politically organized armed groups; (3) violence involving civilians are violent acts by an organized political group against unarmed non-combatants; (4) protests are demonstrations and spontaneous acts of violence by disorganised groups. Columns (5)--(8) include the specific \textit{groups involved}; (6) political militias are armed agents to influence political processes; (7) communal militias are (ethnic or religious) groups engaged in local political competition; (8) rebels are political organizations to counter an established national government. Columns (9)--(10) present results for the specific \textit{dyadic interaction of actors involved}.  Standard errors in parentheses are adjusted for two way clustering by LGA and time with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$. 
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
\end{landscape}
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


% Table 6: Individual Level Perception of Local Governance
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

 \begin{table}[h!]
 \caption{Individual Level Perception of Local Governance \label{table:ab-approval-trust-corrupt}}
 \scalebox{0.7}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{6}{c}}
\toprule
                    &\multicolumn{2}{c}{Approval of LGC Council}                    &\multicolumn{2}{c}{LGC Councillors corrupt}     &\multicolumn{2}{c}{Trust in LGC} \\\cmidrule(lr){2-3}\cmidrule(lr){4-5}\cmidrule(lr){6-7}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)}\\
\midrule
\emph{Panel A:} 6 months \\
\input{tables/fragment-ab6-grievances-ols-yrmo_survey.tex}
\addlinespace
\addlinespace
\emph{Panel B:} 12 months \\
\input{tables/fragment-ab12-grievances-ols-yrmo_survey.tex}
\addlinespace
\addlinespace
\emph{Panel C:} 18 months \\
\input{tables/fragment-ab18-grievances-ols-yrmo_survey.tex}
\addlinespace
\addlinespace


Respondent controls        &          &        X   &          &     X  &           &       X  \\

\bottomrule
\end{tabular}

} 
    \begin{tablenotes} {\footnotesize
\item \textbf{Notes}: All regressions control for time fixed effects and local government area (LGA) fixed effects. The dependent variable in columns (1)--(2) indicates how much a respondent approves of the performance of the local government councillor (4-point Likert scale); in columns (3)--(4) it captures how many of the local councillors a respondent thinks are involved in corruption; columns (5)--(6) indicates how much a respondent trusts the local government council (4-point Likert scale). Panel (A) uses the sum of monthly revenue allocations in the last 6 months, Panel (B) in the last 12 months, and Panel (C) in the last 18 months. The variable \textit{Elected} indicates whether a local government council (LGC) is elected or appointed in a given month. Respondent controls include the respondents age, educational attainment, employment status, gender and an indicator whether the household lives in an urban area. Standard errors in parentheses are clustered by LGA with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$. 
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}

%\end{comment}

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

 % APPENDIX TABLES: ONLINE APPENDIX

%\section*{Tables and Figures for Main Text}
%\begin{comment} 

%\input{APPENDIX_Nigeria_Fetzer_Kyburz_29052018.bbl}

\newpage
\startappendixtables{}
\startappendixfigures{}
 
\appendix
\clearpage

%\section{Additional Tables and Figures \label{sec:appendix_robustness}}
 

\setcounter{footnote}{0} %\setcounter{page}{1} 

\begin{center}
{\LARGE Appendix to ``Cohesive Institutions and Political Violence''}


\bigskip

{\large For Online Publication}

\bigskip

\begin{tabular}[t]{c}
{\large    Thiemo Fetzer  \hspace{2em} Stephan Kyburz }%
\end{tabular}
\\[0pt]
\vspace{1cm} {\large \today}
\end{center}

 
\appendix

\setstretch{1.3}
\section{Additional Figures and Tables}
 

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Figure A1: Conflict around election months
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[h!]
\caption{Conflict around election months}
\begin{center}$
\begin{array}{c}
\includegraphics[scale=.8]{figures/event-nearelection-anyevent-tm-timetoelectime-5.png}
\end{array}$
\begin{minipage}{0.9\textwidth}
\scriptsize{\textbf{Notes}: The figure presents estimated coefficients from a regression with the left hand side being a dummy variable indicating whether there was any conflict event in a local government area (LGA) and month. The regression removes LGA fixed effects and time fixed effects prior and then regresses the residualized dependent variable on a set of dummies capturing the time to the election date. Standard errors are clustered at the LGA level and 10\% confidence bands are indicated.}    
\end{minipage}
\end{center}
 \label{fig:conflictnearelection}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Figure A2: Conventional test for non-linearities
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{landscape}
\begin{figure}[h]
\caption{Conventional test for non-linearities \label{fig:lgcnonlinearities_conventional}}
\begin{center}
% \emph{Panel A:} Events with violence against civilians
$\begin{array}{cc}
%  \includegraphics[scale=.3]<2>{figures/quintile-effect-anymilitaryvspolmilitiaevts-pre.png} &  \includegraphics[scale=.3]<2>{figures/quintile-effect-anymilitaryvspolmilitiaevts-post}  \\
 \includegraphics[scale=.6]{figures/quintile-effect-anypolmilitiavsciviliansevts-pre.png} &  \includegraphics[scale=.6]{figures/quintile-effect-anypolmilitiavsciviliansevts-post}  \\
  
   % \includegraphics[scale=.25]{figures/watercolor-appointed.png} &  \includegraphics[scale=.25]{figures/watercolor-elected.png}  \\


\mbox{(a): Appointed Local Council} &  \mbox{(b): Elected Local Council}\\[3pt]
  \end{array}$

  \end{center}
\scriptsize{\textbf{Notes}: The figures are constructed by demeaning the conflict outcome as well as the gross statutory allocations by the location and time fixed effects, sub-setting the sample into two parts: one with elected and one with appointed local governments. The residuals of the allocations are subdivided into quintiles and we then estimate a simple specification using the quintiles as categorical right hand side measures. The resulting point estimates per quintile are plotted out. The figure displays the effect of local government council (LGC) gross statutory allocations per month on civil conflict by quintile of the shock without (left) and with (right) elected LGC. 90\% confidence intervals obtained from clustering standard errors two way by time and state are indicated.}

\end{figure}
\end{landscape}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Figure A3: Leave-one-out validation: Dropping each state in turn
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{landscape}
\begin{figure}[h]
\caption{Leave-one-out validation: Dropping each state in turn\label{fig:leaveoneout}}
\begin{center}
$\begin{array}{cc}

\mbox{\emph{Panel A:} Reduced form} \\
    \includegraphics[scale=.5]{figures/leave-one-out-main-anyevents.pdf} &  \includegraphics[scale=.5]{figures/leave-one-out-int-anyevents.pdf}  \\
    \mbox{Main effect} &  \mbox{Interaction effect}\\[3pt]
\mbox{\emph{Panel B:} Instrumental variables regression} \\
    \includegraphics[scale=.5]{figures/leave-one-out-iv-main-anyevents.pdf} &  \includegraphics[scale=.5]{figures/leave-one-out-iv-int-anyevents.pdf}  \\
    \mbox{Main effect} &  \mbox{Interaction effect}\\[3pt]

  \end{array}$

  \end{center}
 \scriptsize{\textbf{Notes}: The figure presents results from a leave-one-out exercise. We estimate the main regression including state-by-time and district fixed effects but drop all data pertaining to each of the 37 individual states in turn. The boxplots provide the distribution of the point estimates that are obtained for both the reduced form as well as the instrumental variable estimation.}
\end{figure}

\end{landscape}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

 
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Figure A4: Permutation test on gross statutory allocations and election status
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{figure}[h!t]\centering{
\caption{Permutation test on gross statutory allocations and election status}
\vspace{-0.4cm}
\begin{center}$
\begin{array}{c}
\includegraphics[scale=.8]{figures/permutation-test-interact-both.pdf}
\end{array}$
\end{center}
\vspace{-0.3cm}
\begin{minipage}{0.8\textwidth}
\scriptsize{ \textbf{Notes}: Permutation test on the interaction effect between gross statutory allocations and the elected status dummy. 100 permutations were constructed by randomly reordering spells of appointed vs. elected governments at the local government area (LGA) level. Each model is estimated including state by time and LGA fixed effects. The dashed blue line represents the main effect that is estimated on the LGA Allocations variable, while the red vertical line indicates the estimate that is obtained  on the Elected x LGA Allocations interaction with the true data. The kernel density plots the distribution of the point estimates for the reshuffled Elected x LGA Allocations  interaction variable. It is clear that we can safely reject the null hypothesis with a p-value of less than 0.001.}    
\end{minipage}
 \label{fig:permutation-test}
}
\end{figure}
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Figure A5: Permutation test on election status dummy across 37 states
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{figure}[h!t]\centering{
\caption{Permutation test on election status dummy across 37 states}
\vspace{-0.4cm}
\begin{center}$
\begin{array}{c}
\includegraphics[scale=.8]{figures/permutation-statelevel-test.pdf}
\end{array}$
\end{center}
\vspace{-0.3cm}
\begin{minipage}{0.8\textwidth}
\scriptsize \textbf{Notes}: Permutation test after permuting the election status dummy across the 37 states, thus ignoring part of the local variation. 100 permutations were constructed by randomly reordering spells of appointed vs. elected governments at the state level. Each model is estimated including time and local government area (LGA) fixed effects. The dashed blue line represents the main effect that is estimated on the LGA Allocations variable, while the red vertical line indicates the estimate that is obtained on the Elected x LGA Allocations interaction with the true data. The kernel density plots the distribution of the point estimates for the reshuffled Elected x LGA Allocations  interaction variable. We can reject the null hypothesis of no effect with a p-value of 0.05.
\end{minipage}
   \label{fig:permutation-statelevel-test}
}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Figure A6: Distribution of Ethnic Groups across Nigeria
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[h!t]
\caption{Distribution of Ethnic Groups across Nigeria \label{fig:ethnic-groups-nigeria}}
\centering
\includegraphics[scale=0.5]{figures/NG_map_ethnic_groups_lga.pdf}
\begin{minipage}{0.9\textwidth}
\scriptsize{\textbf{Notes:} The map shows the distribution of ethnic groups across Nigeria. For each local government area (LGA) the largest ethnic group is displayed in a different color. The largest groups are the Yoruba in light blue (South-West), the Hausa-Fulani in pink (North), and the Igbo/Ibo in yellow (South-East) \textit{Sources}: own calculations based on ethnic information in the Demographic Health Surveys (DHS 1990, 2003, 2008, 2010, 2013). Administrative boundaries are from Global Administrative Areas (GADM; \href{https://gadm.org/}{\color{black}https://gadm.org/}).}
\end{minipage}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Figure A7: Population Shares of various ethnic groups in Akwa Ibom and Taraba states
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%\begin{landscape}
 \begin{figure}[h]
 \caption{Population Shares of various ethnic groups in Akwa Ibom and Taraba states \label{fig:pop_shares_hausa-mumuye_taraba}}
\vspace{-0.4cm}
\begin{center}
$\begin{array}{cc}

\text{\emph{Panel A:}  \textbf{Akwa Ibom State}} \\
 \includegraphics[scale=.27]{figures/NG_map_akwa-ibom_ethnicities_Ibibio.pdf} &   \includegraphics[scale=.27]{figures/NG_map_akwa-ibom_ethnicities_oron.pdf} \\
  \mbox{(a) Ibibio} & \mbox{(b) Oron} \\
\\
 \text{\emph{Panel B:}  \textbf{Taraba State}} \\
   \includegraphics[scale=.25]{figures/NG_map_taraba_state_hausa_fulani_pop_share.pdf}  &   \includegraphics[scale=.25]{figures/NG_map_taraba_state_mumuye_pop_share.pdf} \\
 \mbox{(a) Hausa-Fulani} & \mbox{(b) Mumuye}\\

  \end{array}$
\vspace{0.1cm}\\

\begin{minipage}{0.975\textwidth}
\scriptsize{\textbf{Notes:} \textit{Panel A} presents population shares for the (a) Ibibio  and (b) Oron  ethnic groups across local government areas (LGA) in Akwa Ibom State. \textit{Panel B} presents population shares for the (a) Hausa-Fulani  and (b) Mumuye ethnic groups across LGAs in Taraba State. \textit{Sources}: Population shares of ethnic groups across LGAs are calculated based on respondents' information on ethnicity and language in 5 Demographic Health Surveys (DHS 1990, 2003, 2008, 2010, 2013). Administrative boundaries are from Global Administrative Areas (GADM; \href{https://gadm.org/}{\color{black}https://gadm.org/}).}
\end{minipage}
 \end{center}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%\end{landscape}


 \clearpage

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A1: Revenue Allocation Formula
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

 %\clearpage
 %------- Begin LaTeX code -------%

\begin{table}[htbp]\centering \caption{Revenue Allocation Formula \label{table:faacformula}}

 \scalebox{1}{
\centering{
  \begin{threeparttable}
%\caption{LGA Allocations and Conflict: Difference in Difference \label{table:lgc-alloc.tex}}
{

%------- Begin LaTeX code -------%

\begin{tabular}{l c c  }\toprule
\multicolumn{1}{c}{\textbf{Variable}} & \textbf{Mean}
 & \textbf{Std. Dev.} \\ \hline
Equality & 0.138 & 0  \\
Population & 0.104 & 0.058  \\
Internal Revenue Generation effort & 0.009 & 0.004  \\
Landmass & 0.017 & 0.021  \\
Terrain & 0.017 & 0  \\
Health - Hospital Beds & 0.01 & 0.005  \\
Education - Primary enrollment & 0.014 & 0.008  \\
Rain - Water supply spread & 0.005 & 0.006  \\
Rain - Rainfall share & 0.005 & 0  \\
Total index & 0.345 & 0.068  \\
\multicolumn{1}{c}{N} & \multicolumn{2}{c}{774}\\ \bottomrule
\end{tabular}
%------- End LaTeX code -------%
}
\vspace{-0.2cm}
    \begin{tablenotes}
    { \scriptsize {\singlespacing
\item \textbf{Notes}: Mean and Standard deviation of the different sub-indices that feed into the overall index weight used to allocate revenues to local government areas (revenue allocation formula of 2006). \textit{Source}: Federation Account Allocation Committee (FAAC).}
} \end{tablenotes}
      \end{threeparttable}
      
}
}

\end{table}
%------- End LaTeX code -------%
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%



 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A2: Within- and between LGA variation in different types of FAAC Allocations
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

  \begin{table}[h!b]\centering{
 \scalebox{1}{
\centering{
  \begin{threeparttable}
\caption{Within- and between LGA variation in different types of FAAC Allocations \label{table:withinbetweenfaac}}
{

\begin{tabular}{lllll}
\toprule
Variable &  & Mean & Std. Dev. & Observations \\ \midrule

 \\
Total Allocations & overall & 0.305 & 0.243 & N =  140868 \\
 & between &  & 0.074 & n =     774 \\
 & within &  & 0.231 & T =     182 \\
 \\
Statutory Allocations & overall & 0.206 & 0.122 & N =  140868 \\
 & between &  & 0.040 & n =     774 \\
 & within &  & 0.115 & T =     182 \\
 \\
Extraordinary Allocations  & overall & 0.056 & 0.131 & N =  140868 \\
 & between &  & 0.011 & n =     774 \\
 & within &  & 0.131 & T =     182 \\
\bottomrule

\end{tabular}
} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table presents a decomposition of the variation in the Federation Account Allocation Committee (FAAC) allocations within- and between local government areas (LGA). Statutory Allocations are calculated based on a benchmark oil price defined at the beginning of each year. Extraordinary Allocations are additional revenue transfers from the Excess Crude Account (ECA) based on the same allocation formula defined by the FAAC. These allocations are subject to idiosyncratic political decisions. \textit{Source}: information on monthly allocations is published by the FAAC.
} \end{tablenotes}
      \end{threeparttable}
      
}
}
}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A3: Conflict in Oil-Producing States and Oil Fields
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{landscape}
\begin{table}[h!]
\caption{Conflict in Oil-Producing States and Oil Fields \label{table:oil-producing-oil-fields}}
 \scalebox{0.65}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{10}{c}}
\toprule
                &     &\multicolumn{3}{c}{Type of Event }         &\multicolumn{4}{c}{Groups involved}                      &\multicolumn{2}{c}{Between pol militias \&...} \\
                    
                    \cmidrule(lr){3-5}    \cmidrule(lr){6-9} \cmidrule(lr){10-11}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)} &\multicolumn{1}{c}{(9)}&\multicolumn{1}{c}{(10)}\\
                    &\multicolumn{1}{c}{Overall} & \multicolumn{1}{c}{Battle} &\multicolumn{1}{c}{Civilian Violence}&\multicolumn{1}{c}{Protest}&\multicolumn{1}{c}{Military}& \multicolumn{1}{c}{Pol Militia}&\multicolumn{1}{c}{Comm militia}&\multicolumn{1}{c}{Rebels} &\multicolumn{1}{c}{Military}&\multicolumn{1}{c}{Civilians}\\
\midrule

\emph{Panel A:} Oil producing state \\
\input{tables/fragment-sample0-heterogenous-lgc_gsa_mo-oilproducing-anywidelayout-ols-idelec-stm.tex}
\addlinespace
\addlinespace
\emph{Panel B:} LGA with Oil field  \\
\input{tables/fragment-sample0-heterogenous-lgc_gsa_mo-oilgasfielddummy-anywidelayout-ols-idelec-stm.tex}
\addlinespace
\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table presents regression results exploring whether LGAs with oil production activity exhibit a different relationship between revenue shocks and conflict. All regressions control for local government area (LGA) by elected status fixed effects and state by time fixed effects. All dependent variables are binary indicators capturing whether a conflict event occurred in an LGA. The dependent variable includes (1) any violent conflict event defined as the use of force by a group with a political purpose. \textit{Type of events} include (2) battles defined as violent interactions between two politically organized armed groups; (3) violence involving civilians are violent acts by an organized political group against unarmed non-combatants; (4) protests are demonstrations and spontaneous acts of violence by disorganised groups. Columns (5)--(8) include the specific \textit{groups involved}; (6) political militias are armed agents to influence political processes; (7) communal militias are (ethnic or religious) groups engaged in local political competition; (8) rebels are political organizations to counter an established national government. Columns (9)--(10) present results for the specific \textit{dyadic interaction of actors involved}. In Panel (A) the monthly statutory allocation are interacted with a dummy variable indicating whether an LGA is in a state that has any oil-producing facilities; in Panel (B) it is interacted with a dummy variable indicating whether a LGA has any oil field. Standard errors in parentheses are clustered two way by LGA and time with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$. 
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
\end{landscape}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A4: Decomposition of Allocations by Type and Oil Price used
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{landscape} 
 \begin{table}[h!]
 \scalebox{0.85}{
\centering{
  \begin{threeparttable}
\caption{Decomposition of Allocations by Type and Oil Price used \label{table:allocation-decomposition}}
{
\begin{tabular}{l*{9}{c}}
\toprule
                    &\multicolumn{3}{c}{Overall Allocations}         &\multicolumn{3}{c}{Statutory allocations}                      &\multicolumn{3}{c}{Extra allocations} \\
                    
                    \cmidrule(lr){2-4}    \cmidrule(lr){5-7} \cmidrule(lr){8-10}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)}&\multicolumn{1}{c}{(9)}\\
\midrule
\emph{Panel A:} Brent Crude Oil Price \\
\input{tables/allocation-decomposition-brent}
\addlinespace
\addlinespace
\emph{Panel B:} NG Oil production \\
\input{tables/allocation-decomposition-crudeNGA}
\addlinespace

\addlinespace

\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table presents regression results representing a decomposition of the allocation variable using different sets of fixed effects across columns. The dependent variable in columns (1)--(3) is the Overall Allocation determined by the horizontal allocation formula (index) for each local government council (LGC); in columns (4)--(6) the Statutory Allocations calculated based on a benchmark oil price set at the beginning of each year; in columns (7)--(9) the Extraordinary Allocations that are additional revenue transfers from the Excess Crude Account (ECA). The explanatory variable in Panel (A) is an interaction between the total index weight in the allocation formula times the monthly Brent Crude Oil price; in Panel (B) it is an interaction between the total index weight times the monthly Nigerian crude oil production. Standard errors in parentheses are adjusted for two way clustering by time and state-governor with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$. 
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
\end{landscape}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%  RESOURCES AND CONFLICT: ROBUSTNESS CHECKS





%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% ROBUSTNESS POST
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A5: Robustness: Removing election related violence from estimating sample
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{landscape}
 \begin{table}[h!]
 \caption{Robustness: Removing election related violence from estimating sample \label{table:robustness-election}}
 \scalebox{0.65}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{10}{c}}
\toprule
                &     &\multicolumn{3}{c}{Type of Event }         &\multicolumn{4}{c}{Groups involved}                      &\multicolumn{2}{c}{Between Pol. Militias \&...} \\
                    
                    \cmidrule(lr){3-5}    \cmidrule(lr){6-9} \cmidrule(lr){10-11}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)} &\multicolumn{1}{c}{(9)}&\multicolumn{1}{c}{(10)}\\
                    &\multicolumn{1}{c}{Overall} & \multicolumn{1}{c}{Battle} &\multicolumn{1}{c}{Civilian Violence}&\multicolumn{1}{c}{Protest}&\multicolumn{1}{c}{Military}& \multicolumn{1}{c}{Pol. Militia}&\multicolumn{1}{c}{Comm. Militia}&\multicolumn{1}{c}{Rebels} &\multicolumn{1}{c}{Military}&\multicolumn{1}{c}{Civilians}\\
\midrule


\emph{Panel A:} Removing 1 month window around election \\
\input{tables/fragment-int-lgc_gsa_mo-brent-lgc_elec_status-anywidelayout-ols-idelec-stm-sampleelectionrestricted}
\addlinespace
\addlinespace

\emph{Panel B:} Removing 3 month window around election \\
\input{tables/fragment-int-lgc_gsa_mo-brent-lgc_elec_status-anywidelayout-ols-idelec-stm-sampleelectionrestricted2}
\addlinespace
\addlinespace

\emph{Panel C:} Removing 6 month window around election \\
\input{tables/fragment-int-lgc_gsa_mo-brent-lgc_elec_status-anywidelayout-ols-idelec-stm-sampleelectionrestricted3}
\addlinespace
\addlinespace

\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table presents a robustness check removing data around the actual election to document that the changed relationship between resource shocks and conflicts in local government areas (LGA) with elected governments is not confounding any level effects arising due to election related violence.  All regressions control for LGA by elected status and state by time fixed effects. All dependent variables are binary indicators capturing whether a conflict event occurred in an LGA. The dependent variable includes (1) any violent conflict event defined as the use of force by a group with a political purpose. \textit{Type of events} include (2) battles defined as violent interactions between two politically organized armed groups; (3) violence involving civilians are violent acts by an organized political group against unarmed non-combatants; (4) protests are demonstrations and spontaneous acts of violence by disorganised groups. Columns (5)--(8) include the specific \textit{groups involved}; (6) political militias are armed agents to influence political processes; (7) communal militias are (ethnic or religious) groups engaged in local political competition; (8) rebels are political organizations to counter an established national government. Columns (9)--(10) present results for the specific \textit{dyadic interaction of actors involved}. Panel (A) removes the month when a local council election is held from the sample; Panel (B) removes a three months window from the sample; Panel (C) removes a six months window from the sample. \textit{LGC Statutory Allocations} the monthly revenue allocation to a local government council (LGC). The variable \textit{Elected} indicates whether an LGC is elected or appointed in a given month. Standard errors in parentheses are adjusted for two way clustering by time and state-governor with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$. 
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
\end{landscape}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A6: Variance Decomposition of Local Democracy Dummy
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

 \begin{table}[h!]\centering{
 \scalebox{1}{
\centering{
  \begin{threeparttable}
\caption{Variance Decomposition of Local Democracy Dummy \label{table:democracy-dummy-variance-decomposition}}
{
\begin{tabular}{l*{4}{c}}
\toprule
%                                        \cmidrule(lr){2-5} 
                                        
                                                            &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}\\
\midrule
\input{tables/democracy-dummy-variance-decomposition.tex}

\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table presents a decomposition of the variation in the election status dummy variable after controlling for different levels of fixed effects. 
} \end{tablenotes}
      \end{threeparttable}
      
}
}
}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A7: Residual Variation to Hold Local Elections: Allocations and Conflict
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

 \begin{table}[h!]
 \caption{Residual Variation to Hold Local Elections: Allocations and Conflict \label{table:residual_var_elections_allocations}}
 \centering{
 \scalebox{0.7}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{9}{c}}
\toprule
%                                        \cmidrule(lr){2-5} 
                                        
                    &\multicolumn{5}{c}{Allocations }         &\multicolumn{3}{c}{ACLED}                  \\
                    
                    \cmidrule(lr){2-6}    \cmidrule(lr){7-9}        
                                  &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)}&\multicolumn{1}{c}{(9)} \\
\midrule
\input{tables/decomposition-varsexplore1-lgc_elec_status-stm.tex}

\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table presents a series of regressions of the elected status indicator variable (\textit{Elected}) on the different types of allocations for different periods prior to local elections, and conflict events prior to local elections. The residuals in the \textit{Elected} variable are calculated after controlling for state by time fixed effects and local government area (LGA) fixed effects. Column (9) presents the estimation with all relevant explanatory variables included. Standard errors in parentheses are adjusted for two way clustering by time and state-governor with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$. 
} \end{tablenotes}
      \end{threeparttable}
      
}
}
}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A8: Residual Variation in Local Elections: Formula Inputs and Ethnicity
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{landscape}
 \begin{table}[h!]
\caption{Residual Variation in Local Elections: Formula Inputs and Ethnicity \label{table:residual_var_elections_formula}}
 \centering{
 \scalebox{.75}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{111}{c}}
\toprule
%                                        \cmidrule(lr){2-5} 
                                        
                        &\multicolumn{2}{c}{Weather}                      &\multicolumn{5}{c}{Index and Geography} &\multicolumn{3}{c}{Demographics} \\
                    
                    \cmidrule(lr){2-3}  \cmidrule(lr){4-8}\cmidrule(lr){9-11}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)}&\multicolumn{1}{c}{(9)}&\multicolumn{1}{c}{(10)}&\multicolumn{1}{c}{(11)}\\
\midrule
\input{tables/decomposition-varsexplore2-lgc_elec_status-stm.tex}

\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table presents a series of regressions of the elected status indicator variable (\textit{Elected}) on two weather indicators in columns (1)--(2), temperature and annual rainfall, on the allocation index and its sub-indices in columns (3)--(6), on a local government area (LGA) oil field indicator in column (7), and on three ethnic/religious alignment variables in columns (8)--(10), indicating what share of the population is aligned with the state governor's ethnicity and religion, respectively. Column (10) includes all covariates. The residuals in the \textit{Elected} variable are calculated after controlling for state by time fixed effects and LGA fixed effects. Standard errors in parentheses are adjusted for two way clustering by time and state-governor with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$.  
} \end{tablenotes}
      \end{threeparttable}
      
}
}
}
\end{table}
\end{landscape}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%




%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A9: Sensitivity of results to data where there may have been disruption to revenue sharing
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

 \begin{table}[h!]
 \scalebox{0.65}{
\centering{
  \begin{threeparttable}
\caption{Sensitivity of results to data where there may have been disruption to revenue sharing \label{table:int-lgc_gsa_mo-withholding}}
{
\begin{tabular}{l*{10}{c}}
\toprule
                &    \multicolumn{1}{c}{New LGA's}         &\multicolumn{3}{c}{Emergency rule}    &\multicolumn{1}{c}{Non-verified claims}    &\multicolumn{1}{c}{All instances} \\
              \cmidrule(lr){3-5} 
                &    \multicolumn{1}{c}{}         &\multicolumn{1}{c}{Plateau} &\multicolumn{1}{c}{Ekiti}   &\multicolumn{1}{c}{Boko Haram}               \\                    
        %%            \cmidrule(lr){3-5}    \cmidrule(lr){6-9} \cmidrule(lr){10-11}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)} &\multicolumn{1}{c}{(6)} \\
        %            &\multicolumn{1}{c}{Overall} & \multicolumn{1}{c}{Battle} &\multicolumn{1}{c}{Civilian Violence}&\multicolumn{1}{c}{Protest}&\multicolumn{1}{c}{Military}& \multicolumn{1}{c}{Pol. Militia}&\multicolumn{1}{c}{Comm. Militia}&\multicolumn{1}{c}{Rebels} &\multicolumn{1}{c}{Military}&\multicolumn{1}{c}{Civilians}\\
\midrule
\addlinespace
\multicolumn{6}{l}{\emph{Panel A:} Dropping State x Month's which may have been affected by FAAC disruption}  \\
\input{tables/fragment-rob-withholding-month-int-anypoliticalmilitiaevents-lgc_gsa_mo.tex}
\addlinespace
\addlinespace
\multicolumn{6}{l}{\emph{Panel B:} Dropping State x Year's which may have been affected by FAAC disruption}  \\
\input{tables/fragment-rob-withholding-int-anypoliticalmilitiaevents-lgc_gsa_mo.tex}
\addlinespace
\addlinespace
\addlinespace
\multicolumn{6}{l}{\emph{Panel C:} Dropping all data from all states which may have been affected by FAAC disruption}  \\
\input{tables/fragment-rob-withholdingmax-int-anypoliticalmilitiaevents-lgc_gsa_mo.tex}
%\input{tables/fragment-int-lgc_gsa_mo-brent-lgc_elec_status-anywidelayout-iv-idelec-stm-bokoharamstate.tex}

\addlinespace
\addlinespace

\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table presents a robustness check removing data from periods where there \emph{may} have been some disruption to revenue sharing. In Panel A we drop all data for state and month pairs in which there is some indication that transfers may have been disrupted even if they only affected a select few individual local government areas (LGA) within a state. In Panel B we drop all state and year pairs where we know that some LGA by months may have seen some disruption. In Panel C we drop all data for all states for which we ever noted some disruption. The different reasons for disruptions are indicated in the column head and further described in the text. Column (5) drops all data that was affected by disruption irrespective of the underlying reason for disruption. All regressions control for LGA by elected status and state by time fixed effects. The dependent variable is a binary indicator measuring whether an LGA month had a violent event recorded in ACLED. Standard errors in parentheses are adjusted for two way clustering by LGA and time with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$.  
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A10: Robustness: Controlling for State-Governor Specific and Election Status specific LGA Fixed Effects
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{landscape}
\begin{table}[h!]
\caption{Robustness: Controlling for State-Governor Specific and Election Status specific LGA Fixed Effects \label{table:lgcelecstmgov}}
 \scalebox{0.7}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{10}{c}}
\toprule
                &     &\multicolumn{3}{c}{Type of Event }         &\multicolumn{4}{c}{Groups involved}                      &\multicolumn{2}{c}{Between Pol. Militias \&...} \\
                    
                    \cmidrule(lr){3-5}    \cmidrule(lr){6-9} \cmidrule(lr){10-11}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)} &\multicolumn{1}{c}{(9)}&\multicolumn{1}{c}{(10)}\\
                    &\multicolumn{1}{c}{Overall} & \multicolumn{1}{c}{Battle} &\multicolumn{1}{c}{Civilian Violence}&\multicolumn{1}{c}{Protest}&\multicolumn{1}{c}{Military}& \multicolumn{1}{c}{Pol. Militia}&\multicolumn{1}{c}{Comm. Militia}&\multicolumn{1}{c}{Rebels} &\multicolumn{1}{c}{Military}&\multicolumn{1}{c}{Civilians}\\
\midrule

\emph{Panel A:} without elected LGCs \\
\input{tables/fragment-lgc_gsa_mo-lgc_elec_status-anywidelayout-lgcelecstmgov-tm.tex}
\addlinespace
\addlinespace


\emph{Panel B:} with elected LGCs  \\
\input{tables/fragment-int-lgc_gsa_mo-tm-lgc_elec_status-anywidelayout-lgcelecstmgov-tm.tex}
\addlinespace
\addlinespace
\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table documents that results are similar when controlling for a separate set of district fixed effects for each state governor and for whether a local government area (LGA) has an elected or appointed local government council (LGC). This amounts to estimating, on average, more than five different sets of LGA fixed effects and introduces a large number of irrelevant controls understandably inflating the variance. The dependent variable includes (1) any violent conflict event defined as the use of force by a group with a political purpose. \textit{Type of events} include (2) battles defined as violent interactions between two politically organized armed groups; (3) violence involving civilians are violent acts by an organized political group against unarmed non-combatants; (4) protests are demonstrations and spontaneous acts of violence by disorganised groups. Columns (5)--(8) include the specific \textit{groups involved}; (6) political militias are armed agents to influence political processes; (7) communal militias are (ethnic or religious) groups engaged in local political competition; (8) rebels are political organizations to counter an established national government. Columns (9)--(10) present results for the specific \textit{dyadic interaction of actors involved}. Panel (A) presents estimation results for periods with an appointed LGC, and Panel (B) includes periods with an elected LGC. The main explanatory variable is the monthly allocation to a LGC. The variable \textit{Elected} indicates whether an LGC is elected or appointed in a given month. Standard errors in parentheses are adjusted for two way clustering by LGA and time with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$. 
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
\end{landscape}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A11: Dropping all data from states that may have been impacted by the Boko Haram insurgency: Resource rents and conflict – moderating effect of having an elected local government
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{landscape}
 \begin{table}[h!]
 \scalebox{0.72}{
\centering{
  \begin{threeparttable}
\caption{Dropping all data from states that may have been impacted by the Boko Haram insurgency: Resource rents and conflict -- moderating effect of having an elected local government\label{table:int_lgc_gsa_mo-dropbokostates}}
{
\begin{tabular}{l*{10}{c}}
\toprule
                &     &\multicolumn{3}{c}{Type of Event }         &\multicolumn{4}{c}{Groups involved}                      &\multicolumn{2}{c}{Between Pol. Militias \&...} \\
                    
                    \cmidrule(lr){3-5}    \cmidrule(lr){6-9} \cmidrule(lr){10-11}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)} &\multicolumn{1}{c}{(9)}&\multicolumn{1}{c}{(10)}\\
                    &\multicolumn{1}{c}{Overall} & \multicolumn{1}{c}{Battle} &\multicolumn{1}{c}{Civilian Violence}&\multicolumn{1}{c}{Protest}&\multicolumn{1}{c}{Military}& \multicolumn{1}{c}{Pol. Militia}&\multicolumn{1}{c}{Comm. Militia}&\multicolumn{1}{c}{Rebels} &\multicolumn{1}{c}{Military}&\multicolumn{1}{c}{Civilians}\\
\midrule
\multicolumn{5}{l}{\emph{Panel A:} Dropping Yobe, Borno, Adamawa } \\
\input{tables/fragment-int-lgc_alloc_mo-brent-lgc_elec_status-anywidelayout-ols-ID_lgc-stm-bokoharamstate}
\addlinespace
\addlinespace

\multicolumn{5}{l}{\emph{Panel B:} Dropping Yobe, Borno, Adamawa, Niger and Plateau}  \\
\input{tables/fragment-int-lgc_alloc_mo-brent-lgc_elec_status-anywidelayout-ols-ID_lgc-stm-bokoharamstatebroad}


\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table presents regression results suggesting that the observed effects are robust to dropping data pertaining to states that may have been affected by the Boko Haram insurgency. All regressions control for local government area (LGA) by elected status and state by time fixed effects. The dependent variable includes (1) any violent conflict event defined as the use of force by a group with a political purpose. \textit{Type of events} include (2) battles defined as violent interactions between two politically organized armed groups; (3) violence involving civilians are violent acts by an organized political group against unarmed non-combatants; (4) protests are demonstrations and spontaneous acts of violence by disorganised groups. Columns (5)--(8) include the specific \textit{groups involved}; (6) political militias are armed agents to influence political processes; (7) communal militias are (ethnic or religious) groups engaged in local political competition; (8) rebels are political organizations to counter an established national government. Columns (9)--(10) present results for the specific \textit{dyadic interaction of actors involved}. The main explanatory variable is the monthly revenue allocation to a local government council (LGC). The variable \textit{Elected} indicates whether an LGC is elected or appointed in a given month. Standard errors in parentheses are adjusted for two way clustering by LGA and time with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$.  
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
\end{landscape}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A12: LGA Extraordinary Allocations and Conflict
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{landscape}
\begin{table}[h!]
\caption{LGC Extraordinary Allocations and Conflict \label{table:lgc_extra_alloc_mopost}}
 \scalebox{0.75}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{10}{c}}
\toprule
                &     &\multicolumn{3}{c}{Type of Event }         &\multicolumn{4}{c}{Groups involved}                      &\multicolumn{2}{c}{Between Pol. Militias \&...} \\
                    
                    \cmidrule(lr){3-5}    \cmidrule(lr){6-9} \cmidrule(lr){10-11}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)} &\multicolumn{1}{c}{(9)}&\multicolumn{1}{c}{(10)}\\
                    &\multicolumn{1}{c}{Overall} & \multicolumn{1}{c}{Battle} &\multicolumn{1}{c}{Civilian Violence}&\multicolumn{1}{c}{Protest}&\multicolumn{1}{c}{Military}& \multicolumn{1}{c}{Pol. Militia}&\multicolumn{1}{c}{Comm. Militia}&\multicolumn{1}{c}{Rebels} &\multicolumn{1}{c}{Military}&\multicolumn{1}{c}{Civilians}\\
\midrule
\emph{Panel A:} OLS \\
\input{tables/fragment-int-lgc_extra_alloc_mo-brent-lgc_elec_status-anywidelayout-ols-idelec-stm.tex}
\addlinespace
\addlinespace
\emph{Panel B:} IV \\
\input{tables/fragment-int-lgc_extra_alloc_mo-brent-lgc_elec_status-anywidelayout-iv-idelec-stm.tex}

\addlinespace
\addlinespace

\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table focuses on the extraordinary revenue allocations as a shock measure contrasting with the automatically triggered statutory allocations studied in the main paper. All regressions control for local government area (LGA) by elected status fixed effects and state by time fixed effects. The dependent variable includes (1) any violent conflict event defined as the use of force by a group with a political purpose. \textit{Type of events} include (2) battles defined as violent interactions between two politically organized armed groups; (3) violence involving civilians are violent acts by an organized political group against unarmed non-combatants; (4) protests are demonstrations and spontaneous acts of violence by disorganised groups. Columns (5)--(8) include the specific \textit{groups involved}; (6) political militias are armed agents to influence political processes; (7) communal militias are (ethnic or religious) groups engaged in local political competition; (8) rebels are political organizations to counter an established national government. Columns (9)--(10) present results for the specific \textit{dyadic interaction of actors involved}. The variable \textit{LGC Extra Allocations} indicates the extraordinary allocations to each local government council (LGC) from the Excess Crude Account (ECA) on a monthly basis. Extraordinary allocations are based on idiosyncratic political decisions. The variable \textit{Elected} indicates whether an LGC is elected or appointed in a given month. The instrumental variable estimation in Panel (B) uses the index weight interacted with the oil price as instrument. Standard errors in parentheses are adjusted for two way clustering by LGA and time with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$. 
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
\end{landscape}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A13: Effect of Resource rents on civil conflict: Alternative conflict data
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{landscape}
  \begin{table}[h!]
 \caption{Effect of Resource rents on civil conflict: Alternative conflict data\label{table:alternativeconflictdata_post}}
 \scalebox{0.85}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{7}{c}}
\toprule
                    &\multicolumn{3}{c}{ACLED }         &\multicolumn{2}{c}{UCDP GED}                      &\multicolumn{2}{c}{GTD} \\

                    \cmidrule(lr){2-4}    \cmidrule(lr){5-6} \cmidrule(lr){7-8}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}\\
                    &\multicolumn{1}{c}{Events} &\multicolumn{1}{c}{Fatalities}&\multicolumn{1}{c}{Non Boko Haram}&\multicolumn{1}{c}{Events}&\multicolumn{1}{c}{Fatalities}&\multicolumn{1}{c}{Events}&\multicolumn{1}{c}{Fatalities}\\
\midrule
\emph{Panel A:} Any conflict \\
\input{tables/fragment-int-lgc_gsa_mo-tm-lgc_elec_status-anyotherdatalayout-tm.tex}
\addlinespace
\addlinespace


\emph{Panel B:} Levels  \\
\input{tables/fragment-int-lgc_gsa_mo-tm-lgc_elec_status-levotherdatalayout-tm.tex}
\addlinespace
\addlinespace
\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table documents that we find similar results when studying alternative conflict event data. The dependent variable in columns (1)--(3) is based on the ACLED data; in columns (4)--(5) on the Uppsala Conflict Data Program Geo-referenced Event Data, and in columns (6)--(7) on the Global Terrorism Database. All regressions control for time fixed effects and local government area (LGA) fixed effects.  Columns (1), (4), and (6) indicate any conflict event; columns (2), (5), and (7) indicate the number of fatalities in conflict; column (3) only includes conflict events that are \textit{not} associated with the Jihadist militant organization ``Boko Haram'' in northeastern Nigeria. The variable \textit{LGC Statutory Allocations} captures the monthly revenue allocations to each local government council (LGC). The variable \textit{Elected} indicates whether an LGC is elected or appointed in a given month. Panel (A) reports results for any conflict event, and Panel (B) conflict levels. Standard errors in parentheses are adjusted for clustering by LGA with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$. 
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
\end{landscape}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A14: Effect of Resource rents on civil conflict: Different transformations of dependent variables
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{landscape}
\begin{table}[h!]
\caption{Effect of Resource rents on civil conflict: Different transformations of dependent variables \label{table:alternativedependentvariablespost}}
 \scalebox{0.65}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{10}{c}}
\toprule
                &     &\multicolumn{3}{c}{Type of Event }         &\multicolumn{4}{c}{Groups involved}                      &\multicolumn{2}{c}{Between Pol. Militias \&...} \\
                    
                    \cmidrule(lr){3-5}    \cmidrule(lr){6-9} \cmidrule(lr){10-11}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)} &\multicolumn{1}{c}{(9)}&\multicolumn{1}{c}{(10)}\\
                    &\multicolumn{1}{c}{Overall} & \multicolumn{1}{c}{Battle} &\multicolumn{1}{c}{Civilian Violence}&\multicolumn{1}{c}{Protest}&\multicolumn{1}{c}{Military}& \multicolumn{1}{c}{Pol. Militia}&\multicolumn{1}{c}{Comm. Militia}&\multicolumn{1}{c}{Rebels} &\multicolumn{1}{c}{Military}&\multicolumn{1}{c}{Civilians}\\
\midrule
\emph{Panel A:} Any conflict \\
\input{tables/fragment-int-lgc_gsa_mo-year-lgc_elec_status-anywidelayout-ols.tex}
\addlinespace
\addlinespace
\emph{Panel B:} Levels \\
\input{tables/fragment-int-lgc_gsa_mo-year-lgc_elec_status-levelwidelayout-ols.tex}
\addlinespace
\addlinespace
\emph{Panel C:} log(Levels per capita) \\
\input{tables/fragment-int-lgc_gsa_mo-year-lgc_elec_status-logpcwidelayout-ols.tex}
\addlinespace
\addlinespace
\emph{Panel C:} Levels per capita \\
\input{tables/fragment-int-lgc_gsa_mo-year-lgc_elec_status-pcwidelayout-ols.tex}

\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table presents regression results documenting that we find similar results when exploring different transformations of the dependent variable. To study the intensive margin the data is aggregated to the yearly level. All regressions control for local government area (LGA) by time fixed effects. The variable \textit{LGC Statutory Allocations} captures the monthly revenue allocations to each local government council (LGC). The variable \textit{Elected} indicates whether an LGC is elected or appointed in a given month. Standard errors in parentheses are adjusted for clustering by LGA with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$. 
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
\end{landscape}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A15: Different temporal resolution of the data
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{landscape}
 \begin{table}[h!]
 \caption{Different temporal resolution of the data \label{table:alternativetemporalresolutionpost}}
 \scalebox{0.7}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{10}{c}}
\toprule
                &     &\multicolumn{3}{c}{Type of Event }         &\multicolumn{4}{c}{Groups involved}                      &\multicolumn{2}{c}{Between Pol. Militias \&...} \\
                    
                    \cmidrule(lr){3-5}    \cmidrule(lr){6-9} \cmidrule(lr){10-11}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)} &\multicolumn{1}{c}{(9)}&\multicolumn{1}{c}{(10)}\\
                    &\multicolumn{1}{c}{Overall} & \multicolumn{1}{c}{Battle} &\multicolumn{1}{c}{Civilian Violence}&\multicolumn{1}{c}{Protest}&\multicolumn{1}{c}{Military}& \multicolumn{1}{c}{Pol. Militia}&\multicolumn{1}{c}{Comm. Militia}&\multicolumn{1}{c}{Rebels} &\multicolumn{1}{c}{Military}&\multicolumn{1}{c}{Civilians}\\
\midrule

\emph{Panel A:} Annual \\
\input{tables/fragment-int-lgc_gsa_mo-year-lgc_elec_status-anywidelayout-ols}
\addlinespace
\addlinespace
\emph{Panel B:} Quarterly \\
\input{tables/fragment-int-lgc_gsa_mo-tq-lgc_elec_status-anywidelayout-ols.tex}
\addlinespace
\addlinespace
\emph{Panel C:} Monthly \\
\input{tables/fragment-int-lgc_gsa_mo-tm-lgc_elec_status-anywidelayout-ols.tex}

\addlinespace
\addlinespace

\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table presents results documenting that the results are robust to studying different temporal resolutions of the data. Panel A focuses on annual aggregation of the data; panel B focuses on quarterly while Panel C focuses on monthly data. All regressions control for time fixed effects and local government area (LGA) fixed effects. The variable \textit{LGC Statutory Allocations} captures the monthly revenue allocations to each local government council (LGC). The variable \textit{Elected} indicates whether an LGC is elected or appointed in a given month. Panel (A) reports results for an annual resolution of the data, Panel (B) for a quarter yearly resolution, and Panel (C) for a monthly resolution. Standard errors in parentheses are adjusted for clustering by LGA with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$. 
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
\end{landscape}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A16: Effect of Resource rents on civil conflict: Alternative functional forms to account for count data
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

 \begin{table}[h!]
 \caption{Effect of Resource rents on civil conflict: Alternative functional forms to account for count data\label{table:alternativefunctionalformspost}}
   \scalebox{0.62}{
  \begin{threeparttable}
\begin{tabular}{l*{10}{c}}
\toprule
                &     &\multicolumn{3}{c}{Type of Event }         &\multicolumn{4}{c}{Groups involved}                      &\multicolumn{2}{c}{Between pol militias \&...} \\
                    
                    \cmidrule(lr){3-5}    \cmidrule(lr){6-9} \cmidrule(lr){10-11}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)} &\multicolumn{1}{c}{(9)}&\multicolumn{1}{c}{(10)}\\
                    &\multicolumn{1}{c}{Overall} & \multicolumn{1}{c}{Battle} &\multicolumn{1}{c}{Civilian Violence}&\multicolumn{1}{c}{Protest}&\multicolumn{1}{c}{Military}& \multicolumn{1}{c}{Pol Militia}&\multicolumn{1}{c}{Comm militia}&\multicolumn{1}{c}{Rebels} &\multicolumn{1}{c}{Military}&\multicolumn{1}{c}{Civilians}\\
\midrule
\emph{Panel A:} OLS\\
\input{tables/fragment-int-reghdfe-levelwidelayout-lgc_gsa_mo}
\addlinespace
\addlinespace
\emph{Panel B:} Poisson \\
\input{tables/fragment-int-ppmlhdfe-levelwidelayout-lgc_gsa_mo}
\addlinespace
\addlinespace
\emph{Panel C:} NB \\
\input{tables/fragment-int-xtnbreg-levelwidelayout-lgc_gsa_mo}

\addlinespace
\addlinespace

\bottomrule
\end{tabular}

    \begin{tablenotes} 
\item \textbf{Notes}: Table presents regression results documenting that the main results are robust to alternative functional forms. All regressions control for local government area (LGA) and time fixed effects fixed effects. The dependent variables throughout are the count number of conflict events per LGA and year for the period covering 1999-2014. The dependent variable includes the count number of (1) violent conflict events defined as the use of force by a group with a political purpose. \textit{Type of events} include (2) battles defined as violent interactions between two politically organized armed groups; (3) violence involving civilians are violent acts by an organized political group against unarmed non-combatants; (4) protests are demonstrations and spontaneous acts of violence by disorganised groups. Columns (5)--(8) include the specific \textit{groups involved}; (6) political militias are armed agents to influence political processes; (7) communal militias are (ethnic or religious) groups engaged in local political competition; (8) rebels are political organizations to counter an established national government. Columns (9)--(10) present results for the specific \textit{dyadic interaction of actors involved}. The variable \textit{LGC Allocations} captures the monthly revenue allocations to each local government council (LGC). The variable \textit{Elected} indicates whether a LGC is elected or appointed in a given month. Panel (B) employs a Poisson estimation, and Panel (C) a negative-binomial estimation. Standard errors in parentheses are adjusted for clustering by LGA with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$. 
 \end{tablenotes}
      \end{threeparttable}
     
}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%




%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A17: Robustness: Controlling flexibly for formula inputs
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\clearpage
\begin{landscape}
 \begin{table}[h!]
 \caption{Robustness: Controlling flexibly for formula inputs \label{table:formulaflexiblecontrolint_post}}
 \scalebox{0.6}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{10}{c}}
\toprule
                &     &\multicolumn{3}{c}{Type of Event }         &\multicolumn{4}{c}{Groups involved}                      &\multicolumn{2}{c}{Between Pol. Militias \&...} \\
                    
                    \cmidrule(lr){3-5}    \cmidrule(lr){6-9} \cmidrule(lr){10-11}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)} &\multicolumn{1}{c}{(9)}&\multicolumn{1}{c}{(10)}\\
                    &\multicolumn{1}{c}{Overall} & \multicolumn{1}{c}{Battle} &\multicolumn{1}{c}{Civilian Violence}&\multicolumn{1}{c}{Protest}&\multicolumn{1}{c}{Military}& \multicolumn{1}{c}{Pol. Militia}&\multicolumn{1}{c}{Comm. Militia}&\multicolumn{1}{c}{Rebels} &\multicolumn{1}{c}{Military}&\multicolumn{1}{c}{Civilians}\\
\midrule


\emph{Panel A:} Population Weight Decile x Time FE  \\
\input{tables/fragment-robqq_population_wei_tm-int-lgc_gsa_mo-brent-lgc_elec_status-anywidelayout-ols-idelec-stm}
\addlinespace
\addlinespace

\emph{Panel B:} Landmass Weight Decile x Time FE  \\
\input{tables/fragment-robqq_landmass_weigh_tm-int-lgc_gsa_mo-brent-lgc_elec_status-anywidelayout-ols-idelec-stm.tex}
\addlinespace
\addlinespace

\emph{Panel C:} Public Good Access Index Deciles x Time FE  \\
\input{tables/fragment-robqq_primary_enroll_tm-int-lgc_gsa_mo-brent-lgc_elec_status-anywidelayout-ols-idelec-stm.tex}
\addlinespace
\addlinespace

\emph{Panel D:} Water supply spread Index Deciles x Time FE  \\
\input{tables/fragment-robqq_water_supply_s_tm-int-lgc_gsa_mo-brent-lgc_elec_status-anywidelayout-ols-idelec-stm.tex}

\addlinespace
\addlinespace

\emph{Panel E:} Hospital beds Index Deciles x Time FE  \\
\input{tables/fragment-robqq_hospital_beds__tm-int-lgc_gsa_mo-brent-lgc_elec_status-anywidelayout-ols-idelec-stm.tex}


\addlinespace
\addlinespace
\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: Table presents regression results documenting that the results are robust to flexibly controlling for different inputs into the revenue sharing formula by controlling for non-linear time effects in the deciles of the different weight subcomponents. All regressions control local government area (LGA) by elected status and state by time fixed effects. The dependent variable includes (1) any violent conflict event defined as the use of force by a group with a political purpose. \textit{Type of events} include (2) battles defined as violent interactions between two politically organized armed groups; (3) violence involving civilians are violent acts by an organized political group against unarmed non-combatants; (4) protests are demonstrations and spontaneous acts of violence by disorganised groups. Columns (5)--(8) include the specific \textit{groups involved}; (6) political militias are armed agents to influence political processes; (7) communal militias are (ethnic or religious) groups engaged in local political competition; (8) rebels are political organizations to counter an established national government. Columns (9)--(10) present results for the specific \textit{dyadic interaction of actors involved}. The variable \textit{LGC Allocations} captures the monthly revenue allocations to each local government council (LGC). The variable \textit{Elected} indicates whether an LGC is elected or appointed in a given month. Panels (A)--(E) control flexibly for the various allocation formula sub-indices defined by the Federation Account Allocation Committee. Standard errors in parentheses are adjusted for two way clustering by LGA and time with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$. 
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
\end{landscape}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%





%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A18: Validation of measure of exclusion of ethnic groups
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%\begin{landscape}
 \begin{table}[h!]
 \caption{Validation of measure of exclusion of ethnic groups \label{table:validation_exclusion_ethnic-group-family}}
 \scalebox{0.75}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{4}{c}}
\toprule
                    &\multicolumn{4}{c}{Alignment with ethnic group of state governor} \\\cmidrule(lr){2-5}
                    
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)} \\
\midrule
\emph{Panel A:} Own ethnic group treated unfairly (dummy) \\
\input{tables/fragment-ab-SometimesUnfairly-ethnicgrievances-AB_excluded_ethn-ols.tex}
\addlinespace
\addlinespace
\emph{Panel B:}  Own ethnic group treated unfairly \\
\input{tables/fragment-ab-AB_ethngr_treated_unfair-ethnicgrievances-AB_excluded_ethn-ols.tex}
\addlinespace
\addlinespace
State FE        &      X    &    X  &     &                 \\
LGA FE        &           &       &  X    &    X      \\
Time FE        &         &    X  &     &    X       \\

\bottomrule
\end{tabular}

} 
    \begin{tablenotes} {\footnotesize
\item \textbf{Notes}: The explanatory variable measures whether a respondent perceives the own ethnic group treated unfairly on a 4-point Likert scale (Panel A), or at least sometimes (Panel B) to be treated unfairly by the government. The dependent variable indicates whether a respondent is of the same ethnic group as the state governor. Standard errors in parentheses are clustered by LGA with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$.  
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
%\end{landscape}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%




%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A19: LGA Allocations and Grievances towards National Political Institutions
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{table}[h!]
\caption{LGC Allocations and Grievances towards National Political Institutions\label{table:ab-alloc-grievancesnational-12m}}
 \scalebox{0.7}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{6}{c}}
\toprule
                    &\multicolumn{2}{c}{Approval of MP}                    &\multicolumn{2}{c}{National MP's are corrupt}     &\multicolumn{2}{c}{Trust in National Assembly} \\\cmidrule(lr){2-3}\cmidrule(lr){4-5}\cmidrule(lr){6-7}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)}\\
\midrule
\emph{Panel A:} 6 months \\
\input{tables/fragment-ab6-grievancesnational-ols-yrmo_survey.tex}
\addlinespace
\addlinespace
\emph{Panel B:} 12 months \\
\input{tables/fragment-ab12-grievancesnational-ols-yrmo_survey.tex}
\addlinespace
\addlinespace
\emph{Panel C:} 18 months \\
\input{tables/fragment-ab18-grievancesnational-ols-yrmo_survey.tex}
\addlinespace
\addlinespace


Respondent controls        &          &        X   &          &     X  &           &       X  \\

\bottomrule
\end{tabular}

} 
    \begin{tablenotes} {\footnotesize
\item \textbf{Notes}: All regressions control for time fixed effects and local government area (LGA) fixed effects. The dependent variable in columns (1)--(2) indicates how much a respondent approves of the performance of the Members of Parliament at the federal government level; in columns (3)--(4) it captures perception of corruption of Members of Parliament; columns (5)--(6) indicates how much a respondent trusts the National Assembly. The variable \textit{Elected} indicates whether a local government council (LGC) is elected or appointed in a given month. Respondent controls include the respondents age, educational attainment, employment status, gender and an indicator whether the household lives in an urban area. Standard errors in parentheses are clustered by LGA with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$.  
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%




%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A20: LGA Allocations and Grievances towards State level Political institutions
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{table}[h!]
\caption{LGC Allocations and Grievances towards State level Political institutions\label{table:ab-alloc-grievancesstate-12m}}
 \scalebox{0.7}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{6}{c}}
\toprule
                    &\multicolumn{2}{c}{Approval State Governor}                    &\multicolumn{2}{c}{State Assembly corrupt}     &\multicolumn{2}{c}{Trust State governor} \\\cmidrule(lr){2-3}\cmidrule(lr){4-5}\cmidrule(lr){6-7}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)}\\
\midrule
\emph{Panel A:} 6 months \\
\input{tables/fragment-ab6-grievancesstate-ols-yrmo_survey.tex}
\addlinespace
\addlinespace
\emph{Panel B:} 12 months \\
\input{tables/fragment-ab12-grievancesstate-ols-yrmo_survey.tex}
\addlinespace
\addlinespace
\emph{Panel C:} 18 months \\
\input{tables/fragment-ab18-grievancesstate-ols-yrmo_survey.tex}
\addlinespace
\addlinespace


Respondent controls        &          &        X   &          &     X  &           &       X  \\

\bottomrule
\end{tabular}

} 
    \begin{tablenotes} {\footnotesize
\item \textbf{Notes}: All regressions control for time fixed effects and local government area (LGA) fixed effects. The dependent variable in columns (1)--(2) indicates how much a respondent approves of the performance of the State Governor; in columns (3)--(4) it captures perception of corruption among members of the State assembly; columns (5)--(6) indicates how much a respondent trust the state governor. The variable \textit{Elected} indicates whether a local government council (LGC) is elected or appointed in a given month. Respondent controls include the respondents age, educational attainment, employment status, gender and an indicator whether the household lives in an urban area. Standard errors in parentheses are clustered by LGA with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$.  
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%





%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A21: Continuous measure of non-aligned ethnic group share - Inclusion of Ethnic Groups Through Elections and Conflict: Estimation of effect on conflict levels
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{landscape}
  \begin{table}[h!]
  \caption{Continuous measure of non-aligned ethnic group share - Inclusion of Ethnic Groups Through Elections and Conflict: Estimation of effect on conflict levels \label{table:conflict-levels}}
 \scalebox{0.75}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{10}{c}}
\toprule
                &     &\multicolumn{3}{c}{Type of Event }         &\multicolumn{4}{c}{Groups involved}                      &\multicolumn{2}{c}{Between Pol. Militias \&...} \\
                    
                    \cmidrule(lr){3-5}    \cmidrule(lr){6-9} \cmidrule(lr){10-11}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)} &\multicolumn{1}{c}{(9)}&\multicolumn{1}{c}{(10)}\\
                    &\multicolumn{1}{c}{Overall} & \multicolumn{1}{c}{Battle} &\multicolumn{1}{c}{Civilian Violence}&\multicolumn{1}{c}{Protest}&\multicolumn{1}{c}{Military}& \multicolumn{1}{c}{Pol. Militia}&\multicolumn{1}{c}{Comm. Militia}&\multicolumn{1}{c}{Rebels} &\multicolumn{1}{c}{Military}&\multicolumn{1}{c}{Civilians}\\
\midrule
\emph{Panel A:} Level effect  \\
\input{tables/fragment-leveleffectplain-lgc_excl_ethngr_share-lgc_elec_status-anywidelayout-ols-ID_lgc-tm.tex}
\addlinespace
\addlinespace

\multicolumn{3}{l}{\emph{Panel B:} Ethnic alignment heterogenous effect}  \\
\input{tables/fragment-leveleffect-lgc_excl_ethngr_share-lgc_elec_status-anywidelayout-ols-ID_lgc-tm.tex}
\addlinespace
\addlinespace

\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: All regressions control for local government area (LGA) fixed effects and time fixed effects. The dependent variable includes (1) any violent conflict event defined as the use of force by a group with a political purpose. \textit{Type of events} include (2) battles defined as violent interactions between two politically organized armed groups; (3) violence involving civilians are violent acts by an organized political group against unarmed non-combatants; (4) protests are demonstrations and spontaneous acts of violence by disorganised groups. Columns (5)--(8) include the specific \textit{groups involved}; (6) political militias are armed agents to influence political processes; (7) communal militias are (ethnic or religious) groups engaged in local political competition; (8) rebels are political organizations to counter an established national government. Columns (9)--(10) present results for the specific \textit{dyadic interaction of actors involved}. The main explanatory variable \textit{Elected} indicates whether a local government council (LGC) is elected or appointed in a given month. The variable \textit{Non-aligned ethnic group} measures the population share that is of a different ethnic group than the state governor. Standard errors in parentheses are adjusted for two way clustering by LGA and time with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$. 
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
\end{landscape}
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%




%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Table A22: Continuous measure of non-aligned ethnic group share - Inclusion of Ethnic Groups Through Elections and Conflict: Triple Difference-in-Differences with Allocated Rents
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\begin{landscape}
  \begin{table}[h!]
  \caption{Continuous measure of non-aligned ethnic group share - Inclusion of Ethnic Groups Through Elections and Conflict: Triple Difference-in-Differences with Allocated Rents \label{table:triple-diff}}
 \scalebox{0.65}{
\centering{
  \begin{threeparttable}
{
\begin{tabular}{l*{10}{c}}
\toprule
                &     &\multicolumn{3}{c}{Type of Event }         &\multicolumn{4}{c}{Groups involved}                      &\multicolumn{2}{c}{Between Pol. Militias \&...} \\
                    
                    \cmidrule(lr){3-5}    \cmidrule(lr){6-9} \cmidrule(lr){10-11}
                    &\multicolumn{1}{c}{(1)}&\multicolumn{1}{c}{(2)}&\multicolumn{1}{c}{(3)}&\multicolumn{1}{c}{(4)}&\multicolumn{1}{c}{(5)}&\multicolumn{1}{c}{(6)} &\multicolumn{1}{c}{(7)}&\multicolumn{1}{c}{(8)} &\multicolumn{1}{c}{(9)}&\multicolumn{1}{c}{(10)}\\
                    &\multicolumn{1}{c}{Overall} & \multicolumn{1}{c}{Battle} &\multicolumn{1}{c}{Civilian Violence}&\multicolumn{1}{c}{Protest}&\multicolumn{1}{c}{Military}& \multicolumn{1}{c}{Pol. Militia}&\multicolumn{1}{c}{Comm. Militia}&\multicolumn{1}{c}{Rebels} &\multicolumn{1}{c}{Military}&\multicolumn{1}{c}{Civilians}\\
\midrule

\emph{Panel A:} Appointed local governments  \\
\input{tables/fragment-sample0-heterogenous-lgc_gsa_mo-lgc_excl_ethngr_share-anywidelayout-ols-idelec-tm}
\addlinespace
\addlinespace


\addlinespace
\addlinespace


\emph{Panel B:} including interaction terms  \\
\input{tables/fragment-triplediff-lgc_gsa_mo-lgc_excl_ethngr_share-lgc_elec_status-anywidelayout-ols-idelec-tm.tex}


\bottomrule
\end{tabular}

} 
    \begin{tablenotes} 
    {\footnotesize
\item \textbf{Notes}: All regressions control for local government area (LGA) by election status fixed effects and time fixed effects. Panel A is estimated off the sample with appointed local governments and includes LGA fixed effects, while Panel B includes the periods with elected local governments and a separate set of LGA fixed effects for periods in which governments are elected. The dependent variable includes (1) any violent conflict event defined as the use of force by a group with a political purpose. \textit{Type of events} include (2) battles defined as violent interactions between two politically organized armed groups; (3) violence involving civilians are violent acts by an organized political group against unarmed non-combatants; (4) protests are demonstrations and spontaneous acts of violence by disorganised groups. Columns (5)--(8) include the specific \textit{groups involved}; (6) political militias are armed agents to influence political processes; (7) communal militias are (ethnic or religious) groups engaged in local political competition; (8) rebels are political organizations to counter an established national government. Columns (9)--(10) present results for the specific \textit{dyadic interaction of actors involved}. The variable \textit{Elected} indicates whether a local government council (LGC) is elected or appointed in a given month. The variable \textit{Non-aligned ethnic group} measures the population share that is of a different ethnic group than the state governor. \textit{LGC Allocations} captures the monthly revenue allocation to an LGC. Standard errors in parentheses are adjusted for two way clustering by LGA and time with stars indicating *** $p<0.01$, ** $p<0.05$, * $p<0.1$. 
} \end{tablenotes}
      \end{threeparttable}
      
}
}
\end{table}
\end{landscape}
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
%

%
%
%
%
%
%
%
%

%
%
%
%
\clearpage


\end{document}








